You may notice some images loading slow across the Open Collections website. Thank you for your patience as we rebuild the cache to make images load faster.

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays in development economics on gender and tribes in India Maity, Bipasha 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2016_september_maity_bipasha.pdf [ 5.48MB ]
Metadata
JSON: 24-1.0307290.json
JSON-LD: 24-1.0307290-ld.json
RDF/XML (Pretty): 24-1.0307290-rdf.xml
RDF/JSON: 24-1.0307290-rdf.json
Turtle: 24-1.0307290-turtle.txt
N-Triples: 24-1.0307290-rdf-ntriples.txt
Original Record: 24-1.0307290-source.json
Full Text
24-1.0307290-fulltext.txt
Citation
24-1.0307290.ris

Full Text

Essays in Development Economics onGender and Tribes in IndiabyBipasha MaityB.Sc., University of Calcutta, 2008M.S., Indian Statistical Institute, 2010A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Economics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)July 2016c© Bipasha Maity 2016AbstractThis thesis studies the situation of women and tribes in India through theroles of workfare programme, availability of public healthcare and history.The second chapter studies the effect of India’s National Rural Employ-ment Guarantee Programme (NREGA) on consumption expenditure andtime-use, especially on account of women’s participation. Using instrumen-tal variables estimation strategy to deal with the endogeneity in the numberof days worked, we find that women’s participation benefits children, espe-cially girls. Higher spending on nutritious foods, education of girls, lowerengagement of women in domestic chores and greater time spent in schoolfor younger girls are found on account of the programme.The Scheduled Castes (SCs) and Scheduled Tribes (STs) are the twomost disadvantaged social groups in India. The third chapter investigateswhether STs lag behind even the SCs in terms of health, a key developmentindicator which has also remained relatively understudied in the literature.Blinder-Oaxaca decomposition method shows that relative to the lack ofdemand for healthcare from the STs, shortage of supply of health servicesin tribal areas appears to be more important in explaining why STs lagbehind even the SCs in nearly all aspects of women’s and children’s health.The chapter argues that STs need to be studied in isolation from the SCsbecause of different historical reasons for the underdevelopment of these twogroups.The fourth chapter studies the long term implications of historical femaleproperty rights on current development outcomes. Historic patterns of wid-owhood for women is a plausible mechanism through which women becameowners of property. Districts with greater relative female landownership inthe past are found to have lower infant mortality, higher literacy rate, bet-iiAbstractter healthcare for and higher labour force participation of women, greaterreporting of and arrests for crimes committed against women and higherwomen’s autonomy. Greater political representation of women, investmentin public goods and greater economic role played by women in agricultureappear to be possible mechanisms that could explain how female propertyrights during colonial time can have long-term effects.iiiPrefaceChapter 4 of the thesis titled “The Legacy of Female Landlords in India” isan ongoing collaborative work with Prof. Siwan Anderson at the VancouverSchool of Economics, the University of British Columbia. I have been equallyinvolved in all stages of this research project, namely conducting statisticalanalysis of the data as well as writing and editing the manuscript.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Consumption and Time-Use Effects of India’s EmploymentGuarantee and Women’s Participation . . . . . . . . . . . . 52.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Institutional Background of NREGA . . . . . . . . . . . . . 102.2.1 NREGA in India . . . . . . . . . . . . . . . . . . . . 102.2.2 NREGA in Andhra Pradesh . . . . . . . . . . . . . . 122.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.4 Estimation Framework and Empirical Strategy . . . . . . . . 182.4.1 Baseline OLS . . . . . . . . . . . . . . . . . . . . . . 182.4.2 Empirical Strategy: Instrumental Variable . . . . . . 192.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.5.1 Consumption Expenditure Variables . . . . . . . . . . 242.5.2 The Impact of Women’s Participation . . . . . . . . . 26vTable of Contents2.5.3 Time-Use Outcome Variables . . . . . . . . . . . . . . 272.5.4 Major Activity Patterns of Adults . . . . . . . . . . . 272.5.5 Effect on Women’s Major Activity Patterns . . . . . 282.5.6 Major Activity Patterns of Children . . . . . . . . . . 292.5.7 Effect on Time-Use Patterns of Girls . . . . . . . . . 292.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.7 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.8 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Comparing Health Outcomes Across Scheduled Tribes andCastes in India . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553.2.1 Health Outcomes . . . . . . . . . . . . . . . . . . . . 563.2.2 Utilization of Health Services . . . . . . . . . . . . . . 583.2.3 Mechanisms . . . . . . . . . . . . . . . . . . . . . . . 583.2.4 Health Infrastructure under NRHM . . . . . . . . . . 593.3 Empirical Specification . . . . . . . . . . . . . . . . . . . . . 593.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613.4.1 Modern Contraception . . . . . . . . . . . . . . . . . 613.4.2 Antenatal Care, Infant Health, Awareness . . . . . . 623.4.3 Women’s Nutritional Status . . . . . . . . . . . . . . 643.4.4 Children’s Mortality . . . . . . . . . . . . . . . . . . . 643.4.5 Children’s Immunization Status . . . . . . . . . . . . 653.5 Plausible Mechanisms . . . . . . . . . . . . . . . . . . . . . . 663.5.1 Differences in Education . . . . . . . . . . . . . . . . 663.5.2 Differences in Household Amenities . . . . . . . . . . 673.5.3 Differences in Exposure to Media . . . . . . . . . . . 673.5.4 Differences in Women’s Status . . . . . . . . . . . . . 673.5.5 Differences in Medical Care . . . . . . . . . . . . . . . 693.5.6 The Role of Various Factors . . . . . . . . . . . . . . 703.6 The Role of the NRHM . . . . . . . . . . . . . . . . . . . . . 723.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74viTable of Contents3.8 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773.9 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874 The Legacy of Female Landlords in India . . . . . . . . . . . 1024.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1044.2.1 Data on Rent Receivers . . . . . . . . . . . . . . . . . 1044.2.2 Correlates of Historic Female Landownership . . . . . 1054.2.3 Matching Historic Districts with Modern Districts . . 1074.2.4 Data on Other Controls . . . . . . . . . . . . . . . . . 1074.2.5 Data on Outcome Variables . . . . . . . . . . . . . . 1094.3 Empirical Specification . . . . . . . . . . . . . . . . . . . . . 1124.3.1 Empirical Specification: District Level Outcome Vari-ables . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.3.2 Empirical Specification: State Level Outcome Vari-ables . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.3.3 Empirical Specification: Individual Level Outcome Vari-ables . . . . . . . . . . . . . . . . . . . . . . . . . . . 1144.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1154.4.1 Literacy Variables . . . . . . . . . . . . . . . . . . . . 1154.4.2 Infant Mortality and Health Outcomes . . . . . . . . 1164.4.3 Crime . . . . . . . . . . . . . . . . . . . . . . . . . . . 1184.4.4 Women’s Labour Market Participation . . . . . . . . 1194.4.5 Women’s Autonomy . . . . . . . . . . . . . . . . . . . 1204.5 Possible Mechanisms . . . . . . . . . . . . . . . . . . . . . . 1224.5.1 Women Politicians and Public Goods . . . . . . . . . 1224.5.2 Women’s Contribution in Agriculture . . . . . . . . . 1234.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1254.7 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1264.8 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1305 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142viiTable of ContentsAppendicesA Appendix to Chapter 2 . . . . . . . . . . . . . . . . . . . . . . 149A.1 Appendix Tables . . . . . . . . . . . . . . . . . . . . . . . . . 149B Appendix to Chapter 3 . . . . . . . . . . . . . . . . . . . . . . 153B.1 Appendix Tables . . . . . . . . . . . . . . . . . . . . . . . . . 153C Appendix to Chapter 4 . . . . . . . . . . . . . . . . . . . . . . 157C.1 Appendix Tables . . . . . . . . . . . . . . . . . . . . . . . . . 157viiiList of Tables2.1 Share of NREGA Earnings in Household Income . . . . . . . 372.2 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . 382.3 Household Characteristics by Getting NREGA Work On time 392.4 Outcome is Log Per Capita Real Monthly Food Expenditure 402.5 Outcomes are Per Capita Real Spending on Different Foods . 412.6 Outcomes are Per Capita Real Spending on Different Non-Food Items . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422.7 Women Workers: Per Capita Real Spending on Foods . . . . 432.8 Women Workers: Per Capita Real Spending on Non-Food Items 442.9 Outcomes are Major Activity Patterns of Adults . . . . . . . 452.10 Potential Differential Effects: Men vs Women . . . . . . . . . 462.11 Outcomes are Major Activity Patterns of Children . . . . . . 472.12 Potential Differential Effects: Boys vs Girls . . . . . . . . . . 482.13 Potential Differential Effects: Younger vs Older Children . . 492.14 Potential Differential Effects: Time Allocation of a Child ina Typical Day . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . 873.2 Woman Knows/Uses Modern Contraception . . . . . . . . . . 883.3 Woman’s Awareness, Prenatal Care, Infant Health . . . . . . 893.4 Woman is Anaemic, Children’s Mortality . . . . . . . . . . . 903.5 Youngest Child’s Immunization Record . . . . . . . . . . . . . 913.6 Differences in Women’s Education . . . . . . . . . . . . . . . 923.7 Differences in Women’s Exposure to Amenities and Media . . 933.8 Differences in Women’s Status: Work and Autonomy . . . . . 943.9 Differences in Experiences at Health Facility . . . . . . . . . . 95ixList of Tables3.10 Differences in Antenatal Care Received . . . . . . . . . . . . . 963.11 Differences in Medical Care During Delivery . . . . . . . . . . 973.12 Knowledge of Modern Contraception: The Relative Contri-butions of Various Factors . . . . . . . . . . . . . . . . . . . . 983.13 Usage of Modern Contraception: The Relative Contributionsof Various Factors . . . . . . . . . . . . . . . . . . . . . . . . 993.14 Woman Got Tetanus Injection: The Relative Contributionsof Various Factors . . . . . . . . . . . . . . . . . . . . . . . . 1003.15 Woman Got Folic Tablets: The Relative Contributions of Var-ious Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014.1 Descriptive statistics of Explanatory Variables . . . . . . . . . 1304.2 Descriptive statistics of Outcome Variables I . . . . . . . . . . 1314.3 Descriptive statistics of Outcome Variables II . . . . . . . . . 1324.4 Correlates of Historic Female Landownership . . . . . . . . . 1334.5 Literacy Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . 1344.6 Health Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . 1354.7 State Level Reported Crimes and Arrests . . . . . . . . . . . 1364.8 Women’s Labour Market Participation . . . . . . . . . . . . . 1374.9 Measures of Women’s Autonomy . . . . . . . . . . . . . . . . 1374.10 Mechanisms: Political Variables, Villages with Public Goodsand Health Seeking Behaviour . . . . . . . . . . . . . . . . . 1384.11 Mechanisms: Log of Agricultural Yields at the District Level,Rice vs Wheat . . . . . . . . . . . . . . . . . . . . . . . . . . 139A.1 Women’s Labour Market Participation as % of Total Workersin 18 Major Indian States . . . . . . . . . . . . . . . . . . . . 150A.2 Some Evidence of Women’s Control over their NREGA Wages 151A.3 Female Share of Employment in NREGA: % of Total Person-Days . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151A.4 Average NREGA Wage and Casual Wage in Rural India . . . 152B.1 Woman is Underweight . . . . . . . . . . . . . . . . . . . . . 154B.2 Gender Differences in Child Mortality . . . . . . . . . . . . . 155xList of TablesB.3 Gender Composition of a Woman’s Children . . . . . . . . . . 156C.1 District Level Crimes per 1000 population . . . . . . . . . . . 157C.2 District Level Outcomes: Colonial Land Tenure System . . . 158C.3 State Level Outcomes: Colonial Land Tenure System . . . . . 159C.4 Woman’s Labour Market Participation: Rural vs Urban . . . 160C.5 Alternative Measures of Women’s Autonomy . . . . . . . . . 161C.6 Mechanisms: Log of Agricultural Yields at the District Level,Rice vs Wheat . . . . . . . . . . . . . . . . . . . . . . . . . . 162C.7 Alternative Specification of Female Property Rights I . . . . . 163C.8 Alternative Specification of Female Property Rights II . . . . 164C.9 Alternative Specification of Female Property Rights III . . . . 165xiList of Figures2.1 Job Card Proforma: Ministry of Rural Development, Gov-ernment of India . . . . . . . . . . . . . . . . . . . . . . . . . 332.2 Job Card Proforma: Ministry of Rural Development, Gov-ernment of India . . . . . . . . . . . . . . . . . . . . . . . . . 332.3 Job Card Proforma: Ministry of Rural Development, Gov-ernment of India . . . . . . . . . . . . . . . . . . . . . . . . . 342.4 Job Card Proforma: Ministry of Rural Development, Gov-ernment of India . . . . . . . . . . . . . . . . . . . . . . . . . 342.5 Source: Young Lives Survey, Round 3 (2009)-NREGA Daysfor the Working Sample . . . . . . . . . . . . . . . . . . . . . 352.6 Source: Young Lives Survey, Round 3 (2009)-At Most 100NREGA Days for the Working Sample . . . . . . . . . . . . . 352.7 Source: Young Lives Survey, Round 3 (2009)-Share of Foodin Total Monthly Expenditure for Working Sample . . . . . . 362.8 Source: Young Lives Survey, Round 3 (2009)-Log Per CapitaMonthly Food Expenditure in Real 2006 Rupees for WorkingSample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.1 Contraception, DHS 1998-99 & 2005 . . . . . . . . . . . . . . 773.2 Antenatal Care, DHS 1998-99 & 2005 . . . . . . . . . . . . . 773.3 Trends in Delivery Care, DHS 1998-99 & 2005 . . . . . . . . . 783.4 Trends in Delivery Care, DHS 1998-99 & 2005 . . . . . . . . . 783.5 Trends in Children’s Immunization, DHS 1998-99 & 2005 . . 793.6 Trends in Children’s Immunization, DHS 1998-99 & 2005 . . 793.7 Tribal Population by States, Census 2011 . . . . . . . . . . . 803.8 Shortfall in Medical Facilities in Tribal Areas . . . . . . . . . 81xiiList of Figures3.9 Shortfall in Medical Facilities in Tribal Areas . . . . . . . . . 813.10 Shortfall of Doctors/Specialists in Tribal Areas . . . . . . . . 823.11 Shortfall of Doctors/Specialists in Tribal Areas . . . . . . . . 823.12 Shortfall of Other Medical Personnel in Tribal Areas . . . . . 833.13 Shortfall of Other Medical Personnel in Tribal Areas . . . . . 833.14 Regional Differences: Shortfall in Medical Facilities in TribalAreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843.15 Regional Differences: Shortfall in Medical Facilities in TribalAreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843.16 Regional Differences: Shortfall of Doctors in Tribal Areas . . 853.17 Regional Differences: Shortfall of Specialists in Tribal Areas . 853.18 Regional Differences: Shortfall of Pharmacists in Tribal Areas 863.19 Regional Differences: Shortfall of Lab Technicians in TribalAreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.1 Female to Male Rent Receivers, 1921 . . . . . . . . . . . . . . 1264.2 Female to Total Rent Receivers, 1921 . . . . . . . . . . . . . . 1274.3 Land Revenue Systems of British India, (Banerjee & Iyer, 2005)1284.4 Proportion Non-Landlord in British India, (Banerjee & Iyer,2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129xiiiAcknowledgementsI am very grateful to my thesis supervisor, Siwan Anderson, for her supportand advice throughout the programme. She has contributed very greatlyin enhancing my understanding of the subject. I am greatly indebted toAshok Kotwal for his many insightful comments and suggestions which haveimmensely enriched my learning experience. I am also very thankful to himfor his constant support and encouragement. I would like to thank PatrickFrancois and Brian Copeland for their advice. I would also like to express mygreat appreciation for Bharat Ramaswami at the Indian Statistical Institutefor being an outstanding teacher and a source of great support.I am very thankful to Nicole Fortin, Mukesh Eswaran, Joshua Gottlieband Marit Rehavi for their numerous comments and words of encourage-ment. Maureen Chin has been extremely considerate and has provided out-standing administrative support throughout the programme, for which Icannot thank her enough.My time in UBC has been made quite enjoyable by some outstandingfriends and colleagues I have had. I would especially want to thank YawenLiang, Natasha D’Souza, Aruni Mitra and Timea Molnar for the their friend-ship and support.I thank my parents for their support and their almost annual Vancouvervisits that made my summers here greatly enjoyable.I would like to express my gratitude to my grandmother for her love.Lastly, my deep appreciation is for Anup for his friendship and love.xivChapter 1IntroductionThis thesis studies the situations of two sections of the Indian society, namelywomen and the Scheduled Tribes, who face disadvantage on account of theirgender or the social group that they belong to.Gender based discrimination is widespread in India. Social norms aboutgender roles imply that women face impediments in accessing education,healthcare and nutrition, participating in the labour market and have lowerautonomy in intra-household decision making. In this context, Chapters 2and 4 analyse the roles of workfare programme and historical property rightsof women respectively in influencing current development outcomes that canparticularly influence women’s well-being in India.Specifically, Chapter 2 examines the impact of India’s National RuralEmployment Guarantee Programme (NREGA) on outcomes that can poten-tially influence household welfare such as consumption expenditure patterns,food security and individual’s time-use. The chapter uses an instrumentalvariable estimation approach to deal with the potential endogeneity in thenumber of days worked. In particular, the variation created by largely ad-ministrative bottlenecks in providing work to some households but not oth-ers in a village within 15 days of registration is used as an instrument. AsNREGA contains certain provisions that encourage women’s participation,the chapter analyses whether the aforementioned outcomes are affected onaccount of greater women’s participation in the programme. As the genderaspect of the programme has remained largely understudied in the literature,this chapter attempts to make a contribution to that end. It is found thathouseholds that have worked a greater number of days under NREGA spendmore on nutritious foods that can potentially raise nutritional status of chil-dren, such as proteins, dairy, fruits and vegetables. No effect is found on1Chapter 1. Introductionthe spending on adult goods like alcohol and tobacco, which in the contextof rural India are largely consumed by adult males. Households are also lesslikely to face food scarcity due to lack of money indicating better householdfood security. There is also an increased likelihood of spending on clothing,footwear, school uniforms and fees of girls. These effects are largely foundon account of greater women’s participation from the household relative tothat of men. The programme is also found to alter major activity patternsof women with no associated effect for men. In particular, women’s engage-ment in domestic chores is found to decline. Also, younger girls are foundto spend more time in school. Therefore, the chapter finds that workfareprogramme like NREGA has the potential to improve children’s well-beingand most of these findings are due to greater relative participation of womenin the programme.Chapter 4 examines the impact of district-level relative female landown-ership in colonial India on current development outcomes pertaining towomen in modern India. Although the previous literature has studied theimpact of historical institutions (like land tenure systems) on current out-comes, the effect of historical women’s property rights has remained largelyunexplored. District-level variation in landownership of women relative tothat of men in 1921 is taken as an indicator of historical female propertyrights. Historical patterns of widowhood for women and therefore malemortality is a plausible mechanism through which women became owners ofproperty and this seems to be supported by historical accounts on femalelandlords. It is found that greater relative property rights for women inthe past is associated with lower infant mortality and better healthcare forwomen at present. Further, greater literacy, women’s labour market par-ticipation, women’s autonomy, reporting of and arrests for crimes againstwomen are also found in districts that had a higher ratio of female to malelandowners in the past. Greater women’s political representation, invest-ment in public goods and importance of women in agriculture are found tobe plausible mechanisms that could explain how female property rights inthe past could have long-term effects.Strict social stratification of the Hindu society in India meant that cer-2Chapter 1. Introductiontain individuals are disadvantaged because they either occupy the loweststatus in India’s hierarchical caste system or they have historically remainedisolated from the mainstream Hindu society. The former social group are theScheduled Castes (SCs) whereas the latter are the Scheduled Tribes (STs).Despite different historical reasons for their relative socio-economic depri-vation, both STs and SCs have been subjected to similar affirmative actionpolicies by the government to aid their development. These include reserva-tion of seats in state legislatures/ parliament, higher educational institutesand jobs in the public sector.Chapter 3 compares health outcomes of STs with not only the highercastes but also specifically with the SCs. This is to examine how STs fareeven relative to the SCs in terms of health. In contrast to the previous liter-ature which has combined the STs and SCs as a single disadvantaged group,this chapter studies STs and SCs separately. This is motivated by the differ-ent historical reasons for their relative underdevelopment. As the previousliterature has largely focused on educational and occupational mobility be-tween the SC/STs and non-SC/STs, this chapter attempts to contributeto the literature by studying health which is a key development indicatorand has remained largely understudied. The chapter finds that STs performpoorly even relative to SCs in nearly all aspects of women’s and children’shealth such as women’s pre and postnatal heath outcomes, knowledge andusage of modern contraceptives, nutritional status, awareness about dis-eases, child health at birth and children’s immunization coverage. The po-tential mechanisms that can explain these health differences are studied.Differences in the demand for healthcare that can likely be influenced bydifferences in education, exposure to the media, access to basic householdamenities (like drinking water, flush toilet, electricity at home) and women’ssocial status are considered. Availability of health services and professionalmedical personnel are considered as indicators of supply of healthcare. Thechapter finds that as ST women enjoy high social status in their society,poor social status of women that usually impedes access to healthcare can-not explain why ST women have poorer health outcomes relative to their SCcounterparts. Using Blinder-Oaxaca decomposition methodology it is found3Chapter 1. Introductionthat the lack of supply of health services can largely explain the health dif-ferences between the STs and SCs, rather than factors that influence thedemand for healthcare.4Chapter 2Consumption and Time-UseEffects of India’sEmployment Guarantee andWomen’s Participation2.1 IntroductionThe National Rural Employment Guarantee Act (NREGA) was passed bythe Parliament of India in 2005. The Act aims to provide at least 100 daysof employment to rural households who are willing to perform unskilled,manual work in a financial year. The main aim of the Act was to providewage employment to rural households and create durable assets in ruralareas. NREGA costs about 1% of India’s GDP (India’s GDP at currentprices was US $ 1.877 trillion in 2013, World Bank) and covers about 11%of the world’s population (Niehaus and Sukhtankar, 2013). The povertyhead count ratio (at national poverty line) of India was 37.2% in 2005, atthe implementation time of NREGA. Given the sheer scale of the numberof people the Act attempts to cover, it is one of the largest anti-povertyprogrammes in the world (Ravallion, World Bank, 2013). Therefore, it isimportant to understand the welfare implications of this programme on ru-ral households. This chapter uses household data from the state of AndhraPradesh in South India to estimate the causal impact of the number of daysworked under NREGA on a number of outcomes that can potentially influ-ence household welfare. These include food and non-food expenditures at52.1. Introductionthe household level, implications for household food security and individualtime-use.NREGA contains provisions that encourage women’s participation (Kheraand Nayak, (2009)). Therefore, it is particularly important to study howconsumption expenditures on different commodities, household food secu-rity and time-use patterns of household members are likely to be affectedbecause of women’s participation. For instance, the potential increase inwomen’s bargaining power in the household due to NREGA participationmay result in increase in spending on commodities that might benefit chil-dren more relative to adults. Further, whether expenditures on girls arelikely to increase on account of NREGA and women’s participation in itneeds to be investigated. The availability of work under NREGA couldalter major activity patterns of adults and children in the household. Ifwomen are more likely to participate in the programme, it is important tostudy whether women’s engagement in domestic chores decreases. Now asadults participate in the programme, there might be two different effects onchildren. On one hand, greater investment in children’s education can takeplace. On the other hand, it is possible that children may have to take timeoff from education to perform domestic tasks when adults participate in theprogramme. This might be true for NREGA as Khera and Nayak (2009) findthat childcare facilities were not common in NREGA worksites in a largenumber of states. This is in spite of the law that requires such facilitiesto be provided. Taking time away from school to perform domestic choreswould adversely affect children’s welfare. Given social norms about genderroles in rural India, this effect may be more prominent for girls. Thus, it isimportant to investigate how children’s activity patterns and time allocationis affected on account of NREGA by their gender and age groups.This chapter finds that working an additional 10 days by a householdunder NREGA is not only associated with around 8% increase in monthlyper capita real household food expenditure, but also increases spending ondairy, fruits and vegetables, proteins which include fish, meat and eggs-all of which can increase nutritional status of children. Further, we findincreased spending on some luxury goods such as edible oil, sugar, spices62.1. Introductionand beverages like tea/coffee and soft drinks. On the other hand, we do notfind any significant change in the spending on adult goods like alcohol andtobacco products. We also find that households are more likely to spend onclothing and footwear, school uniforms and fees of girls 1. Further, almostall of these findings appear to hold for households that have a larger fractionof women participating in the programme relative to men. Also, this chapterfinds that greater number of days worked by the household in NREGA isassociated with reduction in women’s engagement in domestic chores as theirmajor activity with no associated effect on major activity patterns for men.We find a reduction in leisure for both girls and boys and an increase in thetime spent in school only for younger girls. However, there is no significantchange in the time spent by children on domestic tasks. Although reductionin leisure reduces children’s welfare, it is perhaps comforting to find thatthere has been no significant increase in engagement in domestic chores bychildren, specifically girls, on account of greater adult participation in theprogramme.This chapter analyses the effect of the number of days worked underNREGA on household welfare and it is worth mentioning why it is poten-tially important to study the effect of the programme along the intensivemargin. A number of existing studies on NREGA study the effect of partic-ipation in the programme (that is, along the extensive margin) or the effectof living in a district where NREGA is being implemented (intent-to-treateffect) on a number of outcomes2. However, other studies have indicatedthat the demand for workdays under NREGA is far from falling. Usingmicro-data such as the National Sample Survey (NSS), Dutta et. al. (2012)demonstrate that many households wanted more days of employment thanwhat they were provided with in almost all states of India and Mukhopad-1There is conflicting evidence on the impact of NREGA on agricultural wages in ruralIndia. Studies by Azam (2012) and Imbert and Papp (2015) indicate crowding out ofcasual agricultural labour markets on account of NREGA; thereby contributing to therise in agricultural wages. However, studies by Zimmermann (2015) and Mahajan (2015)find limited impact of NREGA on rural, agricultural labour markets. Therefore, it mightbe difficult to conclude that consumption expenditures of non-participants in NREGA canincrease on account of NREGA raising private casual wages in rural India.2Studies by Bose (2015), Imbert and Papp (2015), Gehrke (mimeo) are some examples.72.1. Introductionhayay (2012) illustrates the same for the state of Rajasthan. The media hasclaimed that there has been demand of increasing the number of workdaysunder NREGA in the state of Tamil Nadu, there has occurred increase inthe number of days under NREGA from 100 to 150 in Rajasthan and theGovernment of India had raised the number of NREGA days to 150 from100 for tribal households throughout the country in February, 2014. Thesestudies and claims in the press, therefore, provide motivation to analyse theeffect of the number of days worked under the programme on household wel-fare, in contrast to the existing studies that focus on the extensive marginor attempt to estimate the intent-to-treat effect of the programme.A few studies have documented the importance of NREGA on householdconsumption spending (Ravi and Engler (2015), Liu and Deininger (2013),Bose (2015)). However, most of these studies have estimated the “intent-to-treat” effect of the programme by comparing households living in districtsthat implemented the programme earlier vis-a-vis those that implemented itlater 3 and none of the studies have focussed on how household consumptionexpenditure and food security is likely to be affected on account of women’sparticipation in the programme. In contrast, this chapter attempts to esti-mate the average treatment effect of NREGA and investigates not only howaggregate consumption spending has changed, but also the different typesof goods on which spending has changed as well as whether such changesare on account of women’s participation in the programme relative to thatof men. To the best of our knowledge, the gender aspect of NREGA hasremained largely understudied in the literature and this chapter attemptsto make a contribution to that end. Also, how children’s and adult’s time isaffected on account of adult participation has remained understudied in thecontext of NREGA 4. This chapter also attempts to study whether major3Difference-in-difference estimation strategy by comparing households living in districtsthat implemented the programme earlier vis-a-vis those that implemented it later, beforeand after the programme implementation, needs to be performed with caution as therolling out of the programme was not random and the programme was implemented inpoorer districts first.4Islam and Sivasankaran (2014) provide intent-to-treat effect estimates of the impactof NREGA on children’s time-use and the authors have provided some correlations, thatare not causal estimates, between the number of days worked under the programme and82.1. Introductionactivity patterns of adults and time allocation of children are affected due tolarger adult participation in the programme, by the gender and age groupsof the household members.In this chapter, the causal impact of NREGA is estimated using instru-mental variables (IV) estimation strategy. This is because the OLS estimatesof the number of days worked under the programme could yield biased esti-mates. Even after controlling for household level characteristics that couldpotentially influence the number of days worked, unobserved individual char-acteristics of household members could influence the total number of days ahousehold works under the programme. OLS results may also overestimatethe spending on cheaper sources of calories like rice and underestimate thespending on nutritious, but more expensive sources of food as householdswith higher spending on cereals and lower spending patterns on more expen-sive foods may also want to work more under the programme. We exploit aprovision under NREGA that specifies that households should be providedwith work within 15 days of registration, in order to construct the instru-ment for the likely endogenous explanatory variable of interest-the number ofdays worked. However, it is important to investigate whether being providedwith work within 15 days of registration (henceforth “on time”) is likely tobe a valid instrument. We provide suggestive evidence that household char-acteristics that are possible indicators of household influence (like belong-ing to political parties, donations to political parties), voting behaviour inelections or exposure to information (such as belonging to self-help groups,cooperative societies, participating in solving community/village problems)in a village are unlikely to influence whether households were provided withwork “on time”. This supports the documentation in the existing literaturethat the proficiency of NREGA’s operation is largely on account of admin-istrative bottlenecks, rather than the characteristics of villagers, who are itspotential beneficiaries (Ravi and Engler (2015), Imbert and Papp (2015)).This chapter is organized as follows: Section 2.2 outlines the institutionalbackground of NREGA and particularly in Andhra Pradesh, as it is the stateon which this study is based; Section 2.3 describes the data used; Sectionchildren’s time-use.92.2. Institutional Background of NREGA2.4 presents the estimation framework and empirical strategy; Section 2.5presents the results and Section 2.6 concludes.2.2 Institutional Background of NREGA2.2.1 NREGA in IndiaThe National Rural Employment Guarantee Act (NREGA) was enactedby the Parliament of India in 2005. The Act aims to provide 100 daysof employment to rural households who are willing to perform unskilled,manual work. NREGA was implemented in a phase-wise manner. Therefore,it was first implemented in 200 poorest districts of the country in 2006,thereafter an additional 130 districts received coverage in 2007 and the Actwas extended to the entire country by early 2008.The Act gave a pivotal role to India’s decentralized elected rural bodies,called the Panchayati Raj in the implementation of the programme. House-holds in a village could apply for a “job card” by submitting a written ororal application to their elected village council, called the Gram Panchayat(henceforth,GP). The GP issues the “job card” to the household free of cost,which is used to record the details of the work received by each adult mem-ber, the number of days of work provided, wages paid, the type of NREGAprojects in which the member worked as well as wages received. Figures2.1-2.3 provide an outline of a typical job card. A household can apply forwork, almost at any time during the year, after receiving the job card. Ap-plications are submitted to the GP and the law mandates that employmentshould be provided within 15 days of registration, failing which householdsare eligible to receive unemployment allowance. The daily unemploymentallowance is mandated to be set at at least one-fourth of the wage rate forthe first thirty days and subsequently at half of the wage rate for the rest ofthe financial year. Figure 2.4 depicts how the job card records the paymentof unemployment allowance for the household. Further, NREGA workersshould receive wages weekly and wage payments should not be delayed be-yond a fortnight. The Act mandates that not more than 40% of the total102.2. Institutional Background of NREGAproject expenditures can be devoted to materials/capital. Therefore, bulkof expenditure for each NREGA project is earmarked for labour wage pay-ments (The Gazette of India, Ministry of Law and Justice, 2005). Also,about 50% of NREGA projects are to be planned and executed by the GP.The projects are to be prepared through consultation with the GP residents.The GP forwards the list of recommended projects to the sub-district pro-gramme officer, who in turn forwards it to the district programme officer forfinal technical and financial approval (Afridi, Iversen and Sharan (2014)).During the 2013-2014 financial year at the all-India level, about 51.7 mil-lion households were allotted work under NREGA (which comprised about99% of the number of households who applied for work according to ad-ministrative data). The Act stipulates that one-third of the workers shouldbe women. Also, rural households are free to choose how 100 days of workare to be allocated among household members and this provides womenin the household the opportunity to participate in the programme. Oneof the most important provisions of the Act is equal wages for men andwomen. This is especially important because women often receive lowerwages in rural labour markets relative to men as Appendix Table A.4 de-picts. Also, the Act requires the provision of childcare facilities in worksiteswhere there are more than five children younger than six years of age andthat works should be predominantly provided within the village. Significantdifferences in women’s participation, however, exists across states- both interms of women’s participation in labour market in general as shown in Ap-pendix Table A.1 as well as in NREGA in particular as Appendix Table A.3depicts. The proportion of total person-days generated during 2013-2014attributable to women workers was around 52%. Dreze and Khera (2009)find from their survey in six north Indian states that 79% women workerscollected their own wages and 69% of them kept their wages earned from theprogramme as shown in Appendix Table A.2. Although an all-India studyis unavailable on the extent to which women NREGA workers control theirearnings, the above mentioned field survey documents that a large fraction ofwomen workers are likely to control their wages earned from the programme.In general, the expenditure on labour comprised 75% of the total spending112.2. Institutional Background of NREGAon NREGA projects 5. However, as reported by the press, unemploymentallowance are not paid out in a large number of states even when prospectiveworkers were not provided with work under the programme within 15 daysof registration6.2.2.2 NREGA in Andhra PradeshThe current study is based on household and individual survey data, calledthe Young Lives Survey (YLS), collected from the south Indian state ofAndhra Pradesh (or AP). AP has been lauded as one of the leading per-formers in the implementation of NREGA in the country (Afridi et. al.(2014)). The state provided employment to 3.4 million households duringthe financial year 2013-2014. During this year, 5,948,234 individuals fromthe state received employment in NREGA projects. Of them, 3,184,172 (or54%) were women workers. The share of NREGA expenditures on wage pay-ments was 72% and the remaining on capital/materials 7. Table 2.1 showsthat around 20% of the household earnings in rural AP can be attributedto NREGA, even if one takes into account earnings from crops and trans-fers. Further, AP is known to conduct regular social audits, in contrast tomost other Indian states. Therefore, the state has often been praised for itsattempt to maintain a high standard of accountability in the execution ofNREGA. Although the Act stipulates that households are entitled to receive100 days of wage employment in a financial year, a number of householdshave worked for more than 100 days under NREGA and this is especiallytrue for AP. We find that during 2013-2014, 687,479 households worked formore than 100 days (MIS Reports available at the NREGA Website of Gov-ernment of India as mentioned in Footnote 9 show the number of householdsworking for more than 100 days in AP).NREGA is implemented in AP through the three-tier elected Panchay-5MGNREGA Public Data Portal at http://www.nrega.nic.in/netnrega/home.aspx6“Unemployment allowance under job scheme paid in only 7 states”, Ruhi Tewari,Livemint, March 31 2010; “No unemployment allowance under NREGA in India”, Au-rangzeb Naqshbandi, OneWorld South Asia, Nov 17, 2009.7MIS Reports at http://www.nrega.ap.gov.in/Nregs/. These computations on AP ex-clude the districts of AP that were transferred to the state of Telengana in 2014.122.3. Dataati Raj system, as in other states. There are three tiers of administrationfor NREGA projects in AP- the district, sub-district or mandal and thevillage, which is the lowest administrative unit. Afridi, Iversen and Sharan(2014) depict the officials at the different tiers of the decentralized villagelevel government who are responsible for implementation of the programme.Village councils or GP in AP are reportedly less powerful relative to GPsin other states, such as Kerala or Rajasthan (Afridi et. al. (2014)). Unlikeother states, the role of the GPs are largely limited to recommending thelist and overseeing the implementation of the projects. Sub-district or man-dal official called the Mandal Parishad Development Officer or the MPDO,assisted by the Assistant Programme Officer (or APO) play a major role insanctioning funds and providing technical approval for NREGA projects inAP. Therefore, the mandal officials play a key role in implementing NREGA,unlike other states where the GPs play a very important role (Mukhopad-hyay (2012); Afridi, Iversen and Sharan (2014); Maiorano (2014)). Now,Imbert and Papp (2015) report the findings of the World Bank (2011) intheir paper where they quote “In practice, very few job card holders for-mally apply for work while the majority tend to wait passively for work tobe provided.” They also note that implementation of the programme de-pends on administrative capacity and political will because administratorshave to deal with a large number of issues, such as preparing a shelf ofwork to be undertaken, getting the shelf approved, deciding the budget forundertaking works- all of which require significant administrative capacity.Afridi et. al. (2014) and Maiorano (2014) also note that NREGA in AP issupply rather than demand driven; that is instead of demand for work fromvillagers influencing the functioning of NREGA (as it was envisaged), theproficiency of NREGA’s functioning largely depends on the quality of theadministrators.2.3 DataThe data used for analysis in this chapter are from Round 3 (2009-2010)of the Young Lives Survey (YLS). The YLS is a child-level panel survey132.3. Dataconducted in the state of undivided AP (AP was bifurcated and some of itsdistricts were transferred to form the new state of Telengana in 2014) andthree rounds of the survey have been conducted till now. Round 1 (2002) islargely not comparable with the Rounds 2 and 3. Round 2 (2007), on theother hand, corresponds to the first year of the implementation of NREGA.However, the survey does not provide detailed information on NREGA inRound 2. It was only in Round 3 detailed information on NREGA has beencollected, namely whether the household was provided with work within15 days of registration, the household was paid wages within 15 days ofcompletion of work , childcare facilities were present in the worksite, singlewomen have been denied employment as well as the number of days eachadult household member has worked under the programme and the reasonsfor not participating in the programme at all. Such detailed information isnot provided in Round 2 of the survey and thus it is difficult to use Round2 for the current analysis.The YLS survey was designed in order to cover three major agro-climaticregions of AP. The survey covers the districts of Anantapur, Kadapa, Karim-nagar, Mahbubnagar, Srikaulam, West Godavari and Hyderabad. The sur-vey covers 20 blocks (mandals) across these 7 districts, of which 15 blocks arerural. The survey has been designed to represent the population of AP. Ofthe 7 districts, Hyderabad, is predominantly urban and NREGA has neverbeen implemented there. Out of the remaining 6 districts, the programmehas been in operation in Anantapur, Kadapa, Karimnagar and Mahbubna-gar from 2006, in Srikaulam from 2007 and in West Godavari from 2008.By the time Round 3 of the survey was conducted, the programme wasoperational in all 6 survey districts in AP.The explanatory variable of interest is the number of days worked byadult household members as well as number of days worked by women andthe ratio of the days worked by women to men in a household under NREGA.We compute this information from the household member level survey onthe number of NREGA days worked and aggregate it to the household level.As we can identify the gender of the household members, we can also com-pute the total number of days worked by women and men separately in142.3. Dataa household. The average NREGA days worked by households is around50 (s.d. 57 days). Figures 2.5 and 2.6 depict the distribution of NREGAdays. Although, we find that there are a few households who have completedmore than 100 days of work from the YLS data, it must be noted that thisis reflective of the administrative data obtained for AP.The working sample includes households in rural area (excluding the dis-trict of Hyderabad altogether), that have not moved since 2007 and was reg-istered under the programme during the last 12 months (and therefore haveat least one adult member who had participated, that is, sought work underthe programme). Firstly, we only include households residing in rural areaand exclude the district of Hyderabad altogether because the programmewas implemented only in the rural areas excluding Hyderabad. Secondly,we restrict the households to include only those that have not changed theirlocation of residence since 2007. This can mitigate the possibilities of selec-tive migration (especially to localities where the programme is likely to bewell implemented)8. Information on whether the household was providedwith work within 15 days of registration or not (the proposed instrumentfor the number of days worked) is available for households that have beenregistered under the programme. About 86% of households were registeredunder NREGA. Therefore, the programme has large coverage in AP. Also,around 63% of households in the working sample report having receivedemployment within 15 days of registration.The outcome variables at the household level correspond to consumptionexpenditures and measures of household food security. The household leveloutcome variables pertaining to consumption expenditures is (log) of real(2006 rupees) per capita food expenditure of the household. Figure 2.7depicts that the share of food in total monthly expenditure is around 60%for most households. Figure 2.8 shows the distribution of log per capita real8Further, around 92% of households that had at least one adult member who hadparticipated in the programme report not having worked under NREGA outside the ge-ographical area administered by their GP. Therefore, it is largely uncommon that somehousehold members migrate in response to NREGA jobs available elsewhere. This is con-sistent with the NREGA’s requirement that employment largely needs to be providedwithin one’s GP.152.3. Datamonthly spending on food. We find that the per capita spending on food isaround Rs. 400 (in 2006 prices) for most households, which is around Rs.2000 (around US $ 30) for an average household of 5 members. Further,real per capita consumption expenditures on a variety of food items areconsidered. These include rice, pulses, proteins(eggs, fish and meat), milk,vegetables and fruits, salt/spices, sugar, edible oils, beverages (tea, coffee,soft drinks), alcohol and tobacco products (like cigarettes). The monthlyfood expenditure is the average monthly spending on food during the pastyear, the recall period for expenditures on different foods is 15 days whilethat for alcohol and tobacco products is 30 days. The expenditure on thesefood items are reported as value of these food items bought and consumedby the household. It might be interesting to know whether buying andconsuming foods is important. Panel A of Table 2.2 reports the mean andstandard deviation of the share of each of the foods bought and consumedout of what is bought, consumed out of own stock and transfers receivedby the household. We find that the mean of the share of foods boughtand consumed ranges from a minimum of 46% (for rice) to a maximum of98% (for alcohol and non-alcoholic beverages). Therefore, focusing on howmuch a household bought and consumed could likely capture a household’soverall consumption expenditure pattern as the share of foods bought andconsumed is a large fraction of how much was bought, consumed out ofown stock and received in transfers. The spending on non-food items arealso considered as outcome variables. They include real per capita medicalexpenditures, clothing and footwear for adults and real aggregate householdspending on festivals. Further, whether the household spends any proportionof its budget on clothing and footwear, school uniforms and fees on the“index” child (child on whom detailed information have been collected in thesurvey) who is a female are also considered as outcome variables. The recallperiod for non-food consumption expenditures and measures of householdfood security is the 12 months preceding the survey.The individual level outcome variables correspond to time-use. The sur-vey asks the most important activity that individuals performed during thelast 12 months. The individuals are then asked the number of days in a162.3. Datamonth and the number of hours in a day this activity was done. As individ-ual’s age is reported in the survey, it is possible to categorise the time-useaccording to adults (18 years of age or older) and children (age is less than18 years). The time spent on different activities are agricultural activities(which include activities such as self-employed in agriculture, earning agri-cultural wages, being annual farm servants and any other agricultural work),non-agricultural activities (that includes individuals who are self-employedin manufacturing, business, services, other non-agricultural pursuits; receivewages for non-agricultural work and who are in regular-salaried employment)and household chores. Also for children, the survey collects additional infor-mation on how many hours per day typically a child spent sleeping, doingdomestic tasks, at school, playing or general leisure. The information onthese detailed child level activities have also been used as outcome variables.The summary statistics on other covariates that are controls in the em-pirical analysis are provided in Panel B of Table 2.2. We find that theaverage household has around 5 members; 25% of households are ScheduledCastes (SCs), 17% are Scheduled Tribes (STs); 48% are Other BackwardClasses (OBCs) and the remaining are non-SC/ST/OBC (which includesupper castes). Also, 98% of the households are Hindus and the remain-ing 2% households are Muslims and Christians. The average landholdingis around 2.5 acres, which is low. About 37% of adult household membersare, on average, literate. Around 99% of households are also found to accessPDS (the public distribution system that sells subsidized food grains). Onan average, 93% household heads are male and 98% household heads live inthe household. The average age of the household head is around 40 years,whereas the average household age is around 26 years. About 48% house-hold members are, on an average, male and 55% households are aware ofsocial audits under NREGA. The average earnings of households excludingNREGA wages is 20,515 and real crop earnings is around 7,317 in real 2006rupees (with large standard deviations).172.4. Estimation Framework and Empirical Strategy2.4 Estimation Framework and EmpiricalStrategy2.4.1 Baseline OLSThe baseline estimation technique used in this analysis is the OLS. Thefollowing estimation equation is used for household level outcome variables:yhv = α+ βNREGAdayshv + γXhv + δd ∗ t+ φv + εhv (2.1)The outcome variable yhv corresponds to household h in community v(which can be thought of as similar to a village). NREGAdays is theexplanatory variable of interest. It is the number of days worked by adultmembers of the household in the programme during the last 12 months. Xhvincludes household level controls that can likely influence consumption ex-penditure patters. We include household size, proportion of literate adults,amount of land owned, age of the household head, average age of householdmembers and age squared and the proportion of males in the household. Wealso include dummy variables to control for whether the household is SC,ST, OBC as well as if it is Hindu, Muslim or Christian, has access to PDS,whether the household head is male, the household head lives in the house-hold and if household members are aware that NREGA is subject to socialaudit. We control for household’s income in real 2006 rupees from majorsources excluding NREGA over the preceding 12 months and crop incomefrom the last agricultural year (results remain unchanged even if we do notcontrol for household’s earnings from major sources and crops). We controlfor years since the programme has been in place in each district in orderto account for district specific administrative learning regarding programmeimplementation (t). φv are village/community fixed effects. εhv is the re-gression disturbance term clustered at the village/community level. Foroutcome variables at the individual level such as time-use, we also includea dummy for whether the individual is female, is literate and age in years.Standard errors are clustered at the household level for these regressions.182.4. Estimation Framework and Empirical Strategy2.4.2 Empirical Strategy: Instrumental VariableThe OLS results may be biased. This is because the number of NREGA daysworked by a household in a year is likely to be endogenous. For example, theOLS results may overestimate the association between spending on cheapersources of calories like rice and pulses; underestimate the association be-tween spending on proteins, vegetables, dairy that are more nutritious butrelatively expensive sources of diet and number of days worked. This islikely because poorer households typically spend more on cheaper sources ofcalories relative to more nutritious, expensive ones and may also choose towork more under NREGA. Other unobserved characteristics of householdmembers may influence the number of days households would want to workand may likely bias OLS results. This motivates us to use the instrumentalvariable (IV) estimation strategy to deal with the potential endogeneity inNREGAdays. We exploit the variation that some households in the villagewere provided with work within 15 days of registration and others were not inorder to identify the effect of the number of days worked in the programmeon the outcome variables of interest. We will argue that this variation islargely on account of administrative reasons and, therefore, potentially ex-ogenous to the household. The proposed instrument is, therefore, a binaryvariable that assumes the value 1 if a household received work within 15days of registration and is 0 otherwise.The empirical specification for IV is as follows, where Zhv is the instru-ment:• First stage:NREGAdayshv = αo + ηZhv + γoXhv + δod ∗ t+ φov + ωhv (2.2)• Second stage:yhv = α1 + β1 ̂NREGAdayshv + γ1Xhv + δ1d ∗ t+ φ1v + υhv (2.3)Clearly, for the instrument to be valid, it should be correlated with192.4. Estimation Framework and Empirical Strategythe endogenous explanatory variable of interest, NREGAdays and satisfythe exclusion restriction. The first stage regression is shown in Panel A ofTable 2.4. We find that the instrument is strongly positively correlated withNREGAdays. In other words, households that received employment within15 days of registration are more likely to work a greater number of daysunder the programme relative to those that did not receive work “on time”.Further, the first stage F-stat on the excluded instrument is around 15 whenwe include all controls (as in column (4) of Table 2.4) and, therefore, theinstrument does not appear to suffer from the weak instruments problem.The issue of obtaining work within 15 days of registration or “on time”needs further exploration and is likely associated with the notion of jobrationing under NREGA. Ravi and Engler (2015), in a study of rural house-holds in Medak district of AP, have found incidence of job rationing underthe programme. Job rationing implies that households were willing to par-ticipate in NREGA, but were not provided with work. However, they donot find evidence to support that job rationing was systematically basedon a household’s socio-economic characteristics, but rather depended on thescattered nature of worksites. That is, either villages did not have enoughworksites or existing worksites did not have enough work to provide. Thiswas because work had not started simultaneously in all villages of the dis-trict, especially during the early phases of the programme (2007). Duttaet. al. (2012) use the NSS (National Sample Survey) data from 2009-2010to understand the incidence of job rationing across different states in India.The authors note that many households were likely rationed in the sense thatthey wanted more days of employment than what they were provided with.Given the limitation imposed by data, the authors focus on rationing underNREGA as implying households not being provided with work at all despitebeing registered. The authors find that administrative data from govern-ment websites report almost no unmet demand for NREGA work. This islargely because state and local governments do not have incentive to reportunmet demand as in that case, unemployment allowance need to be paid outand the cost of which would have to be exclusively borne by the state govern-ments. Dutta et. al. (2012) emphasize the importance of household surveys202.4. Estimation Framework and Empirical Strategyinstead of administrative data to understand the constraints on work sup-ply under NREGA. The authors find little evidence that rationing is biasedagainst the poor and the scheme still appeared to reach out to marginalizedsections of the village, like the SC/ST. Imbert and Papp (2015) mention thatthe process of providing work is complex (a list of works to be undertakenneeds to be prepared, the list needs to be approved, funds are to be allottedand approved etc), all of which require significant administrative capacity.The authors conclude that how well NREGA functions largely depends on“supply side factors” like administrative capacity and political will, ratherthan “demand side factors” like poverty or characteristics of villagers whoare potential beneficiaries of the programme.While the first stage regression shows whether the instrument is relevantor not; the exclusion restriction cannot be tested in the situation of oneendogenous variable and one instrument. The exclusion restriction requiresthat the proposed instrument be orthogonal to the unobserved regression dis-turbance term, conditional on the controls. Although the existing literaturesuggests that administrative bottlenecks are likely to result in job rationingunder NREGA, it is still important to check whether households that wereprovided with work “on time” systematically differed from households thatdid not receive timely provision of work. For instance, it could be that moremotivated households are more likely to receive employment “on time” asthey could be more aware of their rights and bargain for timely work provi-sion. However, we have controlled for household’s awareness that NREGAis subjected to social audits 9 as well as the proportion of literate adults inthe household which could influence household’s awareness about their enti-tlements under NREGA in our regressions as controls. Table 2.3 also showsthat there appears to be no significant difference between households thatreceived and did not receive work “on time” in terms of being members ofvillage-level groups such as self-help groups (SHGs) or cooperative societies,attending frequent meetings of such groups or holding leadership positionsin these groups as membership in these groups can contribute to raising9A system of reviewing the functioning of NREGA to maintain accountability andtransparency of utilizing public funds under the programme.212.4. Estimation Framework and Empirical Strategyawareness about worker rights under NREGA. AP has a history of havingSHGs and given the powerlessness of village councils in AP, the SHGs haveplayed an important role in building awareness about worker’s rights underNREGA (Reddy (2012)). However, Reddy (2012) notes that although SHGshave played an important role in raising awareness levels about NREGAentitlements, households and SHGs were not aware that wage seekers areentitled to compensation if no work was provided within 15 days of regis-tration. Reddy (2012) notes that “This was probably because even thoseresponsible for creation of awareness did not anticipate...failure of provisionof work”. Therefore, it is unlikely that belonging to SHGs could create asystematic bias in which households got work “on time”. Also, unionisingNREGA workers in order to raise their awareness about their entitlementswas started by the AP government only from 2012. Therefore, the argumentthat households with higher motivation or increased awareness are the onesthat are likely to get employment within 15 days of registration is unlikelyto hold during the period of our analysis as Table 2.3 also confirms.Secondly, we might be worried that households that have political con-nections and social capital are more likely to receive employment “on time”.Afridi et. al. (2014) and Maiorano (2014) note the important role playedby a village-level administrative official, the Field Assistant (FA), in theimplementation of NREGA. In general, villagers cannot choose who wouldbe appointed as the FA. However, in practice, the FA could be appointedaccording to the preference of the presiding Member of the Legislative As-sembly (MLA) (Maiorano (2014)) and households with political affiliationsimilar to those of the FA/MLA may be favoured. Table 2.3 shows thatthere is no significant difference between households that got and did notget employment “on time” under NREGA in terms of holding positions ofauthority (including political positions) and the number of years for whichsuch positions were held by household members. Table 2.3 also shows thathouseholds receiving and not receiving work “on time” under NREGA donot significantly differ in terms of engagement of household members in thecommunity, politics or past voting behaviour. For example, no significantdifference is found between households in terms of whether any household222.4. Estimation Framework and Empirical Strategymember talked about community problems with other residents of the com-munity/village, took action to solve problems in the community/village,gave cash or gifts to community groups or political parties, participated inany awareness raising campaigns or protest marches/demonstrations duringthe past three years. Further, the likelihood of voting in national and localelections is high among households and there appears to be no significantdifference between households that received and those that did not receiveemployment “on time” in terms of past voting behaviour from Table 2.3.Thus, political connections or social capital that could be associated withpower/influence in the village/community does not appear to be correlatedwith timely provision of work under NREGA. Further, we might be worriedthat households belonging to certain castes in a village may be systemati-cally favoured vis-a-vis others in being provided with work “on time. Forexample, households belonging to the GP head/FA’s caste or tribe may bemore likely to get work “on time”. Also, as Khosla (2011) notes that the twomajor political parties in AP-Telugu Desam Party (TDP) and the Congressobtain their political allegiance from the OBCs and the SCs respectively,household caste and political affiliation could be interrelated and could likelysystematically bias which households get work “on time” relative to othersin a village. Table 2.3 shows that there are significant differences betweenhouseholds that received and did not receive timely provision of work interms of belonging to SC/ST social groups. However, we have controlled forhousehold caste/tribe affiliation in our regression specifications as householdcaste/tribe can influence our outcome variables of interest, independent ofwhether it can likely be correlated with timely provision of work.Lastly, Table 2.3 also shows that there is no significant difference be-tween households that received and did not receive employment “on time”under NREGA on the basis of their land ownership, earnings from sourcesother than NREGA (such as agricultural wages) and crop earnings; all ofwhich can summarize household wealth or income from sources other thanNREGA. Also, household access to other public programmes does not ap-pear to be associated with timely provision of work under NREGA as accessto PDS is found to be high and does not appear to differ significantly between232.5. Resultshouseholds based on timely provision of work. Also, religion, gender and ageof household head, whether the household head lives in the household, aver-age age of the household and the proportion of males in the household do notappear to be different between households that received and those that didnot receive employment “on time”. Further, we have controlled for each ofthese characteristics in our regression specifications as they can potentiallyinfluence our outcome variables, even if they are not found to be correlatedwith timely provision of work. Also, our sample is restricted to householdswho have not changed their location since 2007 which mitigates the possi-bility of selective migration to villages or blocks where NREGA is likely tobe well-implemented. Further, village fixed effects have been controlled forto account for unobserved village characteristics that could influence timelyprovision of work under NREGA. District specific time-trends have also beencontrolled for in our regression specifications to account for administrativelearning in the implementation of NREGA.Therefore conditional on the controls included in our regression spec-ifications, our proposed instrument-whether a household received employ-ment within 15 days of registration, is unlikely to be correlated with un-observed household characteristics (such as motivation, awareness, havinginfluence/power in the village or community).2.5 ResultsWe present the results of our analysis on consumption expenditure and time-use outcome variables in this section.2.5.1 Consumption Expenditure VariablesTables 2.4-2.6 report the OLS and IV results on consumption expenditurepatterns at the household level.Panel A of Table 2.4 reports the IV results, whereas Panel B reportsthe OLS. Column (4) is the most preferred specification as it includes thefull set of controls. While the number of days worked appear to have no242.5. Resultssignificant effect on per capita monthly spending on food overall from theOLS, we find that working an additional 10 days leads to an increase in (log)per capita monthly food expenditure by around 8% from the IV estimation.This magnitude is close to that found by Ravi and Engler (2015). Table 2.5reports the effect of spending on different kinds of foods. Comparing betweenthe OLS estimates in Panels B,D and the IV estimates in Panels A,C ofTable 2.5 we find that the OLS is likely to overestimate the effect of daysworked on spending on cheaper sources of calories like rice and pulses andlikely underestimates the spending on nutritious, but relatively expensivesources of calories like dairy, vegetables and proteins. From the IV results,we do not find any effect on spending on rice and pulses. On the otherhand, working an additional day increases real per capita spending on milk,proteins, vegetables and fruits by around 1-3% relative to their respectivemeans. This may be because poor households need not necessarily be calorie-deficient, but are more likely to be malnourished. So they may be inclinedto spend more on relatively nutritious items rather than cheaper sources ofcalories like rice, from NREGA income. Further, spending on some itemsconsidered to be “luxurious” for the rural poor like edible oils, sugar, non-alcoholic beverages like tea, coffee and soft drinks are also found to increase.On the other hand, the IV estimates do not show any significant increaseon spending on alcohol and tobacco products that are adult goods and inthe context of rural India, mostly consumed by adult males. Panel E alsoreports implications for household food security. We find that householdsthat worked larger number of days are less likely to face situations of foodscarcity and skipping of meals due to shortage of money; indicating thatfood security has improved for households that worked greater number ofdays under the programme 10.Further, from the IV estimates in Panel A of Table 2.6, we find that percapita medical expenditure, spending on adult footwear and household level10We have also added season fixed effects and month of survey fixed effects in alternativespecifications to account for plausible seasonality in consumption expenditure patters, suchas the occurrence of harvest periods or festivals. The coefficients on spending patternsremain almost unaffected even after the inclusion of season or month of survey fixed effects.252.5. Resultsspending on festivals is higher in those households that have worked a greaternumber of days under the programme. Further, Panel C of Table 2.6 showsthat households that have worked larger number of days under NREGA arealso more likely to spend on clothing, footwear, school uniforms and feesof female children. Specifically, working an additional 10 days under theprogramme appears to increase the likelihood of spending on girl’s clothingby 8%, footwear by 7%, school uniforms by 6% and school fees by 4%.2.5.2 The Impact of Women’s ParticipationTables 2.7 and 2.8 present the results when we consider days worked bywomen alone as well as the number of days worked by women relative tomen as explanatory variables. This is motivated by the fact that NREGAencourages women’s participation, which might in turn potentially increasewomen’s bargaining power within the household through their labour mar-ket participation. In that case, women may be more likely to spend oncommodities that benefit children rather than those that disproportionatelybenefit adults. Further, it might also be interesting to see whether girlsreceive any additional benefits on account of women’s participation in theprogramme.We find from the data that, on average, for every day worked by men;women work 6 days. This indicates that a large proportion of NREGAworkers are women. From Table 2.7 we find that irrespective of whetherwe consider women’s days alone (as in Panels A, C and E ) or the ratio ofwomen’s days to men’s days (as in Panels B, D and F ), the baseline findingson food expenditures and food security from Tables 2.4 and 2.5 continue tohold. In particular, greater number of days worked by women relative tomen in the household under the programme is found to increase real percapita spending on milk, proteins, vegetables and fruits, “luxury” itemssuch as salt, edible oils, sugar and non-alcoholic beverages. Further, we donot find any significant effect on the spending on adult goods like alcoholand tobacco. Household food security situation is also found to improve onaccount of women’s participation in the programme relative to that of men.262.5. ResultsTable 2.8 reports the impact of women’s participation on non-food ex-penditures. We find that greater number of days worked by women resultsin larger expenditures on adult footwear and health care. Further, largernumber of days worked by women is found to increase the likelihood ofspending on female child’s clothing, footwear, school uniforms and schoolfees. However, greater number of days worked by women relative to mendoes not seem to influence medical expenditure, spending during festivals,adult clothing or the likelihood of spending on female children. Therefore,the relative difference in the number of days worked by women to men in thehousehold does not appear to significantly influence the spending on femalechildren.Therefore, greater number of days worked by adults in the householdis found to increase expenditures on nutritious, but more expensive foodswhich can also potentially raise the nutritional status of children and im-prove overall household food security. Importantly, many of these benefitsappear to be on account of women’s relative to men’s participation in theprogramme.2.5.3 Time-Use Outcome VariablesHere we study the effect of NREGA on time-use patterns of adults andchildren. We also analyse the differences in time-use patterns by gender ofindividuals.2.5.4 Major Activity Patterns of AdultsTable 2.9 depicts the effect of NREGA on major activity patterns of adultsin the household. Adults were asked what their major activity was duringthe past 12 months and they were then asked how many days in a typicalmonth and how many hours in a typical day they spent performing that ac-tivity. In particular, Table 2.9 reports the effect of NREGA on the numberof days in a month and the number of hours in a day an adult spent perform-ing the activity. Panel A of Table 2.9 shows that the engagement of adultsin non-agricultural activities as their major activity is found to increase,272.5. Resultswhereas engagement in domestic chores is found to decline in householdswhere adults worked greater number of days under the programme. Some-what analogous findings are obtained from Panel C of Table 2.9. Althoughdays spent per month performing domestic chores as a major activity isfound to decline, the impact on hours spent per day performing domesticchores as a major activity does not appear to be significantly influencedon account of NREGA for adults. This indicates that NREGA providesan employment opportunity that can potentially reduce overall engagementin domestic chores in a month; however this need not necessarily translateinto lowering the number of hours devoted to performing domestic tasks ona daily basis. Therefore, it appears that major activity patterns of adultsare affected on account of NREGA. Overall, engagement in non-agriculturalwork is found to rise and that in domestic chores as major activity is foundto decline.2.5.5 Effect on Women’s Major Activity PatternsTable 2.10 depicts the effect of NREGA on major activity patterns of adultsby gender. We find that major activity patterns of women are found tochange on account of the programme. In particular, women’s engagementin non-agricultural work as their major activity is found to increase andthat in domestic chores is found to decline in households where adults workgreater number of days. However, no significant changes in major activitypatterns is found for men. Therefore, it appears that NREGA affects majoractivity patterns of women, while no such effect can be found for men.These findings potentially indicate that as women are more likely toparticipate in the programme, major activity patterns of women are morelikely to be affected relative to those for men. This can plausibly explainthe finding that women’s engagement in non-agricultural work is found toincrease in households where adults work greater number of days under theprogramme.282.5. Results2.5.6 Major Activity Patterns of ChildrenTable 2.11 shows the impact of NREGA on major activity patterns of chil-dren in the household. It must be noted that children are not legally man-dated to participate in NREGA. Therefore, the number of days worked(which is the explanatory variable of interest) refers to the number of NREGAdays worked by adults in the household. Thus, these tables attempt topresent the effect of adult’s participation in the programme on children’smajor activity patterns.Panel A of Table 2.11 shows that greater number of days worked byadults in the household results in greater engagement of children in agricul-tural work as their major activity. An additional 10 days worked by adultsunder the programme increases the number of days in a typical month chil-dren are engaged in agricultural work as their major activity by around 20%(Panel A of Table 2.11) and raises the number of hours in a typical day spentby children in pursuing agricultural work as their major activity by around9% (Panel C of Table 2.11). On the other hand, no significant impact ofadult participation in the programme is found on children’s engagement innon-agricultural work or domestic chores as their major activities 11.2.5.7 Effect on Time-Use Patterns of GirlsIn this section, we study the effect of adults working in NREGA on majoractivity patterns and time allocation of children by gender and age groups.Table 2.12 presents the results on the effect of adult participation underthe programme on major activity patterns of children by gender. We findthat working an additional 10 days under the programme raises the number11Media reports indicate the government’s willingness to defer the starting time forNREGA work in the morning by an hour for women workers in recognition of the situationthat women workers are also largely responsible for performing household chores. On theother hand, the government often mandates that NREGA work should start early inthe morning and should be over by noon in order to avoid workers getting exposed tointense heat conditions during summer months. Further, around 23% of the householdsreport that a childcare facility was available at their last NREGA worksite. Therefore,flexible working hours under NREGA and the availability of childcare facilities in someNREGA worksites could likely explain why we do not find children substituting for adultsin performing domestic tasks when adults participate under NREGA.292.5. Resultsof days in a typical month and the number of hours in a typical day spentperforming agricultural work as a major activity by 30% (top panel of Table2.12) and 12 % (bottom panel of Table 2.12) respectively for boys. We donot find any significant effect of adult participation in the programme onmajor activity patterns of girls. Therefore, it appears that the results inTable 2.11 on children’s increased engagement in agricultural work as theirmajor activity on account of greater adult participation in NREGA is largelydriven by boys.Table 2.13 shows the effect on major activities of children due to adultparticipation in NREGA by their gender and age groups. We classify chil-dren into younger (those that are 9 years of age or younger) and older cohorts(those that are aged between 10 and 18 years). Panel A presents the resultsfor girls, while Panel B corresponds to boys. We find that greater engage-ment in agricultural work as a major activity because of larger number ofdays worked by adults under the programme largely holds for older boys.No significant effects have been found for girls, irrespective of age groups.As it was found that greater number of days worked under NREGA wasassociated with an increase in women’s engagement in non-agricultural workas their major activity; it might be possible that children are substitutingfor adults in agricultural work, particularly older boys.As there may be heterogeneity in time allocation in a day among childrennot only on the basis of gender but also according to age, we consider theeffect of adults working under the programme on children’s time allocationin a typical day by their gender and age group. Table 2.14 presents theseresults. We consider the effect on younger children (those who are youngerthan 9 years) and older children (those who are aged between 10 and 18years) and by their gender. Panel A reports the results for girls. We findthat the effect of adults working in NREGA primarily affects the time allo-cation of younger children in a typical day. Specifically, the time spent byyounger girls in playing/leisure falls by about 65% and that spent in schoolincreases by around 36% for additional 10 days worked by adults under theprogramme. Panel B reports the results for boys. We find no significanteffect of adults working under NREGA on time allocation of older boys, but302.6. Conclusionthe time devoted to playing/leisure by younger boys appears to decline byaround 68% for additional 10 days worked by adults under the programme.As adults work more under NREGA, particularly women, there couldbe two effects on children’s education. On one hand, with adults workingunder the programme, children’s time spent in school could rise as adultsinvest more on children’s education. As women appear to work more underthe programme, this could also potentially benefit female children. On theother hand, as adults work more under the programme, children may have tosubstitute for adults in performing domestic chores. In particular as womenpredominantly perform domestic chores, with greater participation of adultsand specifically of women in the programme, girls may be induced to performdomestic chores by reducing time spent in school. Therefore, these two likelyopposite effects on children’s education can arise when adults, particularlywomen, participate in the labour market. However, it is reassuring to findthat younger girls are likely to spend more time in school and there is nosignificant effect on the time spent performing domestic tasks by girls onaccount of NREGA.2.6 ConclusionIn this chapter, we have studied the effect of the number of days workedby a household under India’s National Rural Employment Guarantee Act(NREGA) on consumption expenditure patterns and individual time-usethrough instrumental variable estimation strategy. As NREGA particularlyencourages the participation of women, we find increased spending on nu-tritious items such as milk, proteins, vegetables and fruits along with some“luxury” food items such as salt/spices, sugar and edible oils as well asbeverages like tea/coffee and soft drinks; but no effect on the spending onadult goods such as tobacco and alcohol. Also, the household food securitysituation is found to improve on account of greater number of days workedunder the programme. Further, we find that households that work a greaternumber of days under NREGA are more likely to spend on clothing andfootwear, school uniforms and fees of girls. Most of these findings are be-312.6. Conclusioncause of greater women’s participation in the programme, relative to men.These findings are consistent with the literature that women and men havedistinct preferences and are likely to spend their incomes on different com-modities, with women being more likely to spend on goods that can raisechildren’s welfare. In terms of time-use outcomes, we find a reduction inthe engagement of women in domestic chores as their major activity and anincrease in time spent performing non-agricultural work, in households thatwork greater number of days under NREGA; with no associated effect formen. However, these findings for adult women do not appear to translateinto greater time spent performing chores for female children, contrary toour expectations. But, we find an increase in engagement in agriculturalwork as a major activity for male children. Leisure/play time is found to belower for both girls and boys in households where adults work greater num-ber of days under NREGA. On the other hand, time spent in school is foundto increase for younger girls. Reassuringly, time spent performing domestictasks by girls is not found to increase on account of the programme.322.7. Figures2.7 FiguresFigure 2.1: Job Card Proforma: Ministry of Rural Development, Govern-ment of IndiaFigure 2.2: Job Card Proforma: Ministry of Rural Development, Govern-ment of India332.7. FiguresFigure 2.3: Job Card Proforma: Ministry of Rural Development, Govern-ment of IndiaFigure 2.4: Job Card Proforma: Ministry of Rural Development, Govern-ment of India342.7. Figures0.005.01.015Density0 100 200 300 400NREGS Total DaysFigure 2.5: Source: Young Lives Survey, Round 3 (2009)-NREGA Days forthe Working Sample0.005.01.015.02.025Density0 20 40 60 80 100NREGS Total DaysFigure 2.6: Source: Young Lives Survey, Round 3 (2009)-At Most 100NREGA Days for the Working Sample352.7. Figures0123Density0 .2 .4 .6 .8 1Share of Food Expenditure in Monthly Total Expenditurekernel = epanechnikov, bandwidth = 0.0299Kernel density estimateFigure 2.7: Source: Young Lives Survey, Round 3 (2009)-Share of Food inTotal Monthly Expenditure for Working Sample0.2.4.6.8Density4 5 6 7 8 9Log Per Capita Monthly Food Expenditure (in real 2006 rupees)kernel = epanechnikov, bandwidth = 0.0976Kernel density estimateFigure 2.8: Source: Young Lives Survey, Round 3 (2009)-Log Per CapitaMonthly Food Expenditure in Real 2006 Rupees for Working Sample362.8. Tables2.8 TablesTable 2.1: Share of NREGA Earnings in Household IncomeVariable Mean Std. Dev. ObservationsNo Crop Income:Sale of Livestock Products 0.07 0.17 1409Agricultural Wages 0.39 0.32 1409Regular Wages/Salary 0.13 0.27 1409Casual Wages 0.15 0.26 1409NREGA Wages 0.25 0.25 1409Selling Commodities 0.02 0.13 1409With Crop Income:Sale of Livestock Products 0.05 0.13 1411Agricultural Wages 0.33 0.30 1411Regular Wages/Salary 0.12 0.26 1411Casual Wages 0.13 0.25 1411NREGA Wages 0.19 0.20 1411Selling Commodities 0.02 0.11 1411Crop Income 0.17 0.25 1411With Transfers:Sale of Livestock Products 0.05 0.12 1413Agricultural Wages 0.31 0.26 1413Regular Wages/Salary 0.11 0.25 1413Casual Wages 0.12 0.23 1413NREGA Wages 0.18 0.18 1413Selling Commodities 0.02 0.11 1413Transfers 0.20 0.16 1413Note: Data source is the Round 3 of the Young Lives Survey (2009-2010). All observationsare at the household level. Share of income computed on the basis of earnings duringthe reference period of last 12 months. Transfers include social subsidy, interest on bankaccount and those from friends/relatives not belonging to the household. The mean shareneed not exactly add up to 1 on account of rounding. Sample contains rural households(excluding district of Hyderabad) who have not moved since 2007 and for whom numberof days worked under NREGA was available.372.8. TablesTable 2.2: Descriptive StatisticsPanel A:Variable Mean Std. Dev. ObservationsShare of Rice Bought 0.46 0.41 1410Share of Pulses Bought 0.66 0.37 1354Share of Milk Bought 0.80 0.40 1242Share of Proteins Bought 0.97 0.14 1003Share of Vegetables, Fruits Bought 0.95 0.17 1411Share of Salt/Spices, Oil, Sugar Bought 0.80 0.21 1413Share of Tea/Coffee/Pop Bought 0.98 0.16 1145Share of Alcohol Bought 0.98 0.11 628Panel B:Variable Mean Std. Dev. ObservationsHousehold Size 5.39 2.08 1413If Scheduled Caste 0.25 0.43 1413If Scheduled Tribe 0.17 0.38 1413If Other Backward Class 0.48 0.50 1413If Hindu 0.98 0.15 1413If Muslim 0.01 0.12 1413If Christian 0.01 0.09 1413Land Owned 2.52 16.20 1413Proportion of Literate Adults 0.37 0.35 1413If Access PDS 0.99 0.09 1413Male Household Head 0.93 0.25 1413Age of Household Head 40.31 9.49 1413If Head Lives in HH 0.98 0.15 1413Average Household Age 26.21 6.04 1413Household Age Squared 723.50 332.79 1413Proportion of Males 0.48 0.15 1413Knows Social Audit 0.55 0.50 1413Earnings (no NREGA) 20515.79 26453.97 1413Crop Earnings 7317.57 29874.56 1413Note: Data source is the Round 3 of the Young Lives Survey (2009-2010). All observations are at thehousehold level. Share of each food item is computed as a share of item bought and consumed from whatis bought, consumed out of own stock and transfers received by the household. The variables “HouseholdSize” “Land Owned”, “Proportion of Literate Adults”, “ Age of Household Head”, “Average HouseholdAge”, “ Household Age Squared”, “Proportion of Males”, “Earnings (no NREGA)” and “Crop Earnings”are continuous variables; all other variables are binary variables that assume the value of 1 if the descriptionprovided is true and assume the value of 0 otherwise. “Land Owned” is reported in acres. Age is reportedin years. “Earnings (no NREGA)” is household earnings from sources listed in Table 2.1, except fromNREGA, in real 2006 rupees earned during the household’s reference period of the last 12 months. “CropEarnings” is household earnings from crops in real 2006 rupees earned during the last agricultural year(2008-2009).382.8. TablesTable 2.3: Household Characteristics by Getting NREGA Work On timeVariable Did Not Get Got Work EquivalenceWork on Time on Time of MeansHousehold Size 5.45 (0.100) 5.34 (0.06) 0.11 (0.11)Access to PDS 0.99 (0.004) 0.99 (0.003) 0.001 (0.005)Land Owned 2.13 (0.15) 2.74 (0.66) -0.61 (0.86)Earnings (no NREGA) 21047.65 (1377.21) 19922.38 (719.06) 1125.27 (1412.52)Crop Earnings 6119.65 (649.04) 8177.74 (1165.73) -2058.09 (1597.37)Scheduled Caste 0.29 (0.02) 0.23 (0.01) 0.06*** (0.02)Scheduled Tribe 0.08 (0.01) 0.22 (0.01) -0.14*** (0.02)Other Backward Classes 0.50 (0.02) 0.48 (0.02) 0.02 (0.03)Hindu 0.97 (0.01) 0.98 (0.004) -0.004 (0.008)Muslim 0.02 (0.01) 0.01 (0.003) 0.01 (0.01)Christian 0.004 (0.003) 0.01 (0.003) -0.007 (0.005)Male Household Head 0.94 (0.01) 0.93 (0.01) 0.01 (0.01)Age of Household Head 40.74 (0.41) 40.25 (0.31) 0.48 (0.51)Head Lives in Household 0.98 (0.01) 0.98 (0.005) 0.002 (0.01)Average Household Age 26.47 (0.26) 26.09 (0.19) 0.38 (0.32)Proportion of Males 0.48 (0.01) 0.48 (0.005) 0.0003 (0.01)Proportion of Literate Adults 0.42 (0.02) 0.35 (0.01) 0.07*** (0.02)If Position Held 0.03 (0.01) 0.03 (0.01) 0.002 (0.01)Number of Years Position Held 0.12 (0.05) 0.14 (0.05) -0.02 (0.07)Member of a Group 0.59 (0.02) 0.59 (0.02) -0.001 (0.03)Group Leader 0.10 (0.01) 0.12 (0.01) -0.02 (0.02)Attended Frequent Meetings 0.58 (0.02) 0.58 (0.02) 0.001 (0.03)Talked about problems 0.25 (0.02) 0.28 (0.01) -0.03 (0.02)Voted in national elections 0.98 (0.01) 0.98 (0.004) -0.005 (0.01)Voted in local elections 0.98 (0.01) 0.99 (0.004) -0.01 (0.01)Gave cash/gifts to groups 0.06 (0.01) 0.06 (0.01) -0.01 (0.01)Taken action about problem 0.17 (0.02) 0.19 (0.01) -0.02 (0.02)Participated in awareness campaigns 0.15 (0.02) 0.17 (0.01) -0.02 (0.02)Participated in protest march 0.07 (0.01) 0.07 (0.01) -0.002 (0.01)Observations 543 919Note: Sample is restricted to include only those households that are in rural area of undivided Andhra Pradesh (excludingHyderabad), have not moved since 2007 and have sought work under NREGA. The variables such as positions held (polit-ical/apolitical), group membership (self-help groups, cooperative societies), group leadership, attending frequent meetingsare from Round 2 of the Young Lives Survey (2007). Standard errors are in parentheses. For all other variables, the datasource is Round 3 of the Young Lives Survey (2009-10). ***, ** and * indicate statistical significance at the 1% , 5% and10% level of significance respectively.392.8. TablesTable 2.4: Outcome is Log Per Capita Real Monthly Food ExpenditurePanel A: IVVariable (1) (2) (3) (4)Number of NREGA Days 0.006*** 0.009*** 0.007** 0.008**(0.002) (0.003) (0.003) (0.003)Mean of Dependent Variable 5.91 5.91 5.91 5.91First Stage:Dep Variable: Number of NREGA DaysIf Got Work on Time 17.52*** 16.71*** 12.18*** 11.65***(3.03) (2.98) (3.07) (3.05)F-Stat on excluded instrument 30.52 29.02 15.88 15.33Observations 1412 1411 1411 1411Panel B: OLSVariable (1) (2) (3) (4)Number of NREGA Days 0.0005* 0.0005* 0.0004 0.0004(0.00026) (0.00027) (0.00025) (0.00025)R-Squared 0.109 0.152 0.360 0.363Controls:Caste Controls Yes Yes Yes YesReligion Controls Yes Yes Yes YesLand Owned Yes Yes Yes YesIf Access PDS Yes Yes Yes YesHH Size Yes Yes Yes YesProportion of Literate Adults Yes Yes Yes YesIf Male HH Head No Yes Yes YesIf Head Lives in HH No Yes Yes YesHH Head Age No Yes Yes YesAverage HH Age No Yes Yes YesAverage HH Age Squared No Yes Yes YesProportion of Males No Yes Yes YesHH Knows Social Audit No Yes Yes YesHH Net Real Income (non-NREGS) No No No YesHH Net Real Crop Income No No No YesEarly Districts*t No Yes Yes YesVillage Fixed Effects No No Yes YesNote: Data source is the Round 3 of the Young Lives Survey (2009-2010). All observations are at the household level.Robust standard errors clustered at the community level are in parentheses. ***, ** and * indicate statistical significanceat the 1% , 5% and 10% level of significance respectively.“Caste Controls” include dummy variable for Scheduled Caste,Scheduled Tribe, Other Backward Classes; “Religion Controls” include dummy variables for Hindu, Muslim, Christian.“Early Districts*t” is the district-specific number of years since NREGA was operational in that district with reference to2008-09. Early districts include Anantapur, Kadapa, Karimnagar, Mahbubnagar, Srikakulam. The district of West Godavarifirst implemented NREGA during 2008-09.402.8.TablesTable 2.5: Outcomes are Per Capita Real Spending on Different FoodsPanel A: IV:Variable Rice Pulses Milk Proteins Veg-FruitsNREGA Days 0.028 0.052 0.171** 0.323* 0.412***(0.162) (0.056) (0.073) (0.170) (0.152)Obs 1410 1411 1408 1393 1409Mean of Dep Var 22.60 10.15 7.41 24.48 30.14Panel B: OLS:Variable Rice Pulses Milk Proteins Veg-FruitsNREGA Days 0.038** 0.004 -0.002 -0.016 -0.003(0.017) (0.005) (0.005) (0.011) (0.009)R-Squared 0.209 0.212 0.181 0.189 0.401Obs 1411 1412 1409 1394 1410Panel C: IV:Variable Sugar/Oil Beverages Alcohol TobaccoNREGA Days 0.302*** 0.176** -0.038 -0.024(0.116) (0.074) (0.177) (0.102)Obs 1411 1402 1399 1410Mean of Dep Var 24.87 5.70 18.41 15.72Panel D: OLS:Variable Sugar/Oil Beverages Alcohol TobaccoNREGA Days -0.001 -0.002 -0.029* 0.004(0.008) (0.005) (0.015) (0.011)R-Squared 0.346 0.229 0.154 0.169Obs 1412 1403 1400 1411Panel E: Implicationsfor HouseholdFood SecurityVariable No Food Often No Food Lower Meals Often Lower MealsNREGA Days -0.002** -0.001** -0.001 -0.003**(0.001) (0.0006) (0.0013) (0.001)Obs 1411 1411 1411 1411Mean of Dep Var 0.04 0.02 0.08 0.04Note: Data source is the Round 3 of the Young Lives Survey (2009-2010). Each cell represents a separate regression. “Proteins”include fish,meat and eggs; “Beverages” include tea/coffee and soft drinks. Expenditure on all food items in real 2006 rupees.Food security questions are binary variables that assume value 1 if the variable description is true and 0 otherwise. “No Food” isfood scarcity due to lack of money, “Often No food” is frequent food scarcity due to lack of money, “Lower Meals” is reducing thenumber of meals taken, “Often Lower Meals” is frequent lowering of number of meals. All observations are at the household level.“Obs” is number of observations and “Mean of Dep Var” is mean of the dependent variable. Robust standard errors clustered atthe community level are in parentheses. ***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significancerespectively. Regression specification is as in Column(4) of Table 2.4.412.8. TablesTable 2.6: Outcomes are Per Capita Real Spending on Different Non-FoodItemsPanel A: IV:Variable Adult Clothing Adult Footwear Medical FestivalsNREGA Days 1.640 0.596** 20.37** 38.06*(1.855) (0.283) (10.17) (20.44)Obs 1411 1367 1407 1411Mean of Dep Var 381.47 59.66 659.68 2214.99Panel B: OLS:Variable Adult Clothing Adult Footwear Medical FestivalsNREGA Days 0.154 0.001 -0.310 0.560(0.281) (0.026) (1.022) (0.840)R-Squared 0.310 0.376 0.120 0.420Obs 1368 1369 1408 1412Panel C: IV:Variable: Girl Child Clothing Footwear School Uniform School FeesNREGA Days 0.008** 0.007** 0.006** 0.004*(0.003) (0.003) (0.003) (0.002)Obs 1411 1411 1411 1411Mean of Dep Var 0.47 0.46 0.27 0.27Panel D: OLS:Variable: Girl Child Clothing Footwear School Uniform School FeesNREGA Days 0.0001 -0.00002 0.0003 0.0002(0.0002) (0.0002) (0.0002) (0.0002)R-Squared 0.337 0.309 0.264 0.184Obs 1412 1412 1412 1412Note: Data source is the Round 3 of the Young Lives Survey (2009-2010). Each cell represents a separate regression. Alloutcome variables on non-food expenditures are expressed in per capita spending in real 2006 rupees, except on festivalswhich is at the household level. “Girl Child Clothing/Footwear/School Uniform/School Fees” is a binary variable whichassumes the value 1 if the household has spent any proportion of its budget on clothing/footwear/school uniform/school feeson the index child, who is a female and is 0 otherwise. All other variables are continuous variables. All observations areat the household level. Robust standard errors clustered at the community level are in parentheses. ***, ** and * indicatestatistical significance at the 1% , 5% and 10% level of significance respectively. Regression specification is as in Column(4)of Table 2.4.422.8.TablesTable 2.7: Women Workers: Per Capita Real Spending on FoodsPanel A:Variable Rice Pulses Milk Proteins Veg-FruitsWomen’s Days -0.002 0.134 0.344** 0.623* 0.837***(0.349) (0.126) (0.073) (0.335) (0.332)Obs 1351 1352 1349 1335 1350Panel B:Variable Rice Pulses Milk Proteins Veg-FruitsWomen’s/Men’s Days -0.188 0.104 0.561* 0.962* 1.349**(0.675) (0.215) (0.299) (0.579) (0.531)Obs 1153 1154 1151 1139 1152Panel C:Variable Sugar/Oil Beverages Alcohol Tobacco Log FoodWomen’s Days 0.644*** 0.371** 0.005 -0.103 0.015**(0.279) (0.132) (0.423) (0.235) (0.008)Obs 1352 1343 1341 1351 1352Panel D:Variable Sugar/Oil Beverages Alcohol Tobacco Log FoodWomen’s/Men’s Days 1.308** 0.678** -0.155 -0.034 0.025**(0.578) (0.306) (0.830) (0.507) (0.012)Obs 1154 1145 1145 1153 1154Panel E:Variable No Food Often No Food Lower Meals Often Lower MealsWomen’s Days -0.005** -0.003** -0.004 -0.007**(0.002) (0.001) (0.003) (0.003)Obs 1352 1352 1352 1352Panel F:Variable No Food Often No Food Lower Meals Often Lower MealsWomen’s/Men’s Days -0.009* -0.006* -0.008 -0.014**(0.005) (0.004) (0.006) (0.006)Obs 1154 1154 1154 1154Note: See Table Notes of Table 2.5.432.8. TablesTable 2.8: Women Workers: Per Capita Real Spending on Non-Food ItemsPanel A:Variable Adult Clothing Adult Footwear Medical FestivalsWomen’s Days 2.439 1.151* 43.22* 78.95(4.091) (0.676) (23.60) (49.23)Observations 1310 1311 1348 1352Panel B:Variable Adult Clothing Adult Footwear Medical FestivalsWomen’s Days/Men’s Days 8.088 2.539* 82.05 154.6(7.851) (1.375) (52.19) (95.70)Observations 1154 1154 1154 1154Panel C:Variable: Girl Child Clothing Footwear School Uniform School FeesWomen’s Days 0.016* 0.014* 0.012* 0.009*(0.009) (0.008) (0.007) (0.006)Observations 1352 1352 1352 1352Panel D:Variable: Girl Child Clothing Footwear School Uniform School FeesWomen’s Days/Men’s Days 0.022 0.020 0.010 0.012(0.016) (0.016) (0.012) (0.010)Observations 1154 1154 1154 1154Note: See Table Notes of Table 2.6.442.8. TablesTable 2.9: Outcomes are Major Activity Patterns of AdultsDays Per MonthPanel A: IV ResultsVariable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.013 0.068** -0.032**(0.028) (0.029) (0.015)Obs 3931 3931 3931Mean of Dep Var 9.03 4.09 1.40Panel B: OLSVariable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.004 0.004 -0.002(0.004) (0.004) (0.002)R-Squared 0.155 0.148 0.105Obs 3934 3934 3934Hours Per DayPanel C: IV:Variable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.001 0.022** -0.004(0.010) (0.010) (0.003)Obs 3931 3931 3931Mean of Dep Var 3.39 1.48 0.27Panel D: OLSVariable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.002 0.002 -0.001(0.001) (0.001) (0.0004)R-Squared 0.170 0.153 0.094Obs 3934 3934 3934Note: Each cell represents a separate regression. Individual’s reference period for work activities is basedon 12 month recall period. Robust standard errors clustered at the household level are in parentheses.***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significance respectively.Regression specification is Column (4) of Table 2.4 and controls for age, dummy for if literate and female.452.8. TablesTable 2.10: Potential Differential Effects: Men vs WomenDays Per MonthIV Results: Women MenVariable Agriculture Work Agriculture WorkNREGA Days 0.040 -0.005(0.034) (0.033)Mean of Dep Var 8.96 9.10Variable Non-Agriculture Work Non-Agriculture WorkNREGA Days 0.077** 0.056(0.033) (0.036)Mean of Dep.Var 2.70 5.60Variable Domestic Chores Domestic ChoresNREGA Days -0.070** 0.003(0.032) (0.002)Mean of Dep Var 2.63 0.07Observations 2047 1884Hours Per DayIV Results Women MenVariable Agriculture Work Agriculture WorkNREGA Days 0.017 -0.013(0.012) (0.012)Mean of Dep.Var 3.45 3.36Variable Non-Agriculture Work Non-Agriculture WorkNREGA Days 0.026** 0.018(0.011) (0.012)Mean of Dep.Var 0.98 2.01Variable Domestic Chores Domestic ChoresNREGA Days -0.009 0.001(0.007) (0.0004)Mean of Dep.Var 0.52 0.01Observations 2047 1884Note: See Table Notes of Table 2.9 for variable definitions, data sources, controls included and other details.462.8. TablesTable 2.11: Outcomes are Major Activity Patterns of ChildrenDays Per MonthPanel A: IV:Variable Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.022* 0.007 -0.006(0.013) (0.008) (0.011)Obs 3835 3835 3835Panel B:OLSVariable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.001 -0.002** 0.003**(0.002) (0.001) (0.001)R-Squared 0.117 0.108 0.065Obs 3838 3838 3838Hours Per DayPanel C: IV:Variable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.009* 0.001 -0.001(0.005) (0.003) (0.003)Obs 3835 3835 3835Panel D:OLS:Variable: Agriculture Non-Agriculture Domestic ChoresNREGA Days 0.0003 -0.001** 0.001*(0.001) (0.0003) (0.0003)R-Squared 0.120 0.110 0.058Obs 3838 3838 3838Note: Each cell represents a separate regression. Individual’s reference period for work activities is basedon 12 month recall period. Robust standard errors clustered at the household level are in parentheses.***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significance respectively.Regression specification is Column (4) of Table 2.4 and controls for age, dummy for if literate and female.472.8. TablesTable 2.12: Potential Differential Effects: Boys vs GirlsDays Per Month:IV Results Female Children Male ChildrenVariable Agriculture Work Agriculture WorkNREGA Days 0.015 0.032**(0.018) (0.016)Variable Non-Agriculture Work Non-Agriculture WorkNREGA Days 0.003 0.008(0.009) (0.013)Variable Domestic Chores Domestic ChoresNREGA Days -0.005 -0.004(0.017) (0.008)Observations 1991 1844Hours Per Day:IV Results Female Children Male ChildrenVariable Agriculture Work Agriculture WorkNREGA Days 0.007 0.012**(0.006) (0.006)Variable Non-Agriculture Work Non-Agriculture WorkNREGA Days -0.001 0.001(0.004) (0.004)Variable Domestic Chores Domestic ChoresNREGA Days 0.0004 -0.002(0.004) (0.002)Observations 1991 1844Note: See Table Notes of Table 2.11 for variable definitions, data sources, controls included and otherdetails.482.8. TablesTable 2.13: Potential Differential Effects: Younger vs Older ChildrenDays Per Days Per Hours Per Hours PerMonth Month Day DayPanel A: GirlsIV Results Younger Older Younger OlderAgriculture:NREGA Days -0.003 0.020 0.0002 0.008(0.008) (0.034) (0.003) (0.012)Non-Agriculture:NREGA Days -0.002 0.014 -0.001 -0.0002(0.002) (0.019) (0.001) (0.007)Domestic Chores:NREGA Days 0.00002 0.007 0.000002 0.004(0.0002) (0.035) (0.00002) (0.008)Observations 950 1041 950 1041Days Per Days Per Hours Per Hours PerMonth Month Day DayPanel B: BoysIV Results Younger Older Younger OlderAgriculture:NREGA Days -0.001 0.116* -0.0004 0.043*(0.004) (0.064) (0.002) (0.024)Non-Agriculture:NREGA Days 0.001 0.021 0.001 0.002(0.002) (0.044) (0.001) (0.014)Domestic Chores:NREGA Days -0.004 0.001 -0.001 -0.004(0.004) (0.024) (0.001) (0.007)Observations 984 860 984 860Note: See Table Notes of Table 2.11 for variable definitions, data sources, controlsincluded and other details. Younger children are aged 9 years or less; older childrenare between ages of 10-18 years.492.8. TablesTable 2.14: Potential Differential Effects: Time Allocation of a Child in aTypical DayPanel A: Female ChildrenIV Results Younger Children Older ChildrenVariable Sleeping SleepingNREGA Days -0.004 -0.001(0.007) (0.005)Mean of Dep Var 9.27 8.65Variable Domestic Task Domestic TaskNREGA Days 0.009 0.001(0.005) (0.007)Mean of Dep Var 0.38 1.35Variable School SchoolNREGA Days 0.036** -0.008(0.017) (0.015)Mean of Dep Var 7.23 6.36Variable Playing/Leisure Playing/LeisureNREGA Days -0.065** 0.006(0.027) (0.013)Mean of Dep Var 5.24 4.29Observations 731 663Panel B: Male ChildrenIV Results Younger Children Older ChildrenVariable Sleeping SleepingNREGA Days 0.018 -0.017(0.011) (0.012)Mean of Dep Var 9.27 8.64Variable Domestic Task Domestic TaskNREGA Days 0.0004 -0.008(0.004) (0.012)Mean of Dep Var 0.23 0.65Variable School SchoolNREGA Days 0.025 -0.016(0.016) (0.025)Mean of Dep Var 7.42 6.86Variable Playing/Leisure Playing/LeisureNREGA Days -0.068** 0.002(0.034) (0.019)Mean of Dep Var 5.17 4.62Observations 758 545Note: The outcome variables refer to number of hours in a day a child does a specificactivity. Younger children are aged 9 years or less; older children are between ages of10-18 years. See Table Notes of Table 2.11 for data sources, controls included and otherdetails.50Chapter 3Comparing Health OutcomesAcross Scheduled Tribes andCastes in India3.1 IntroductionThe Scheduled Tribes (STs) and the Scheduled Castes (SCs) have been thetwo most disadvantaged social groups in India. The SCs occupy the low-est status in the hereditary and hierarchical caste system of India and havebeen subjected to oppression and untouchability by the upper castes 12. Onthe other hand, historically the STs have remained largely isolated fromthe larger Hindu society; often living in remote, hilly and forested areas.Recognizing the socio-economic deprivation that these social groups facedon account of untouchability and isolation, certain provisions were madein the Indian Constitution to aid their development. In fact, Xaxa (2001)notes that there are more provisions for the STs than for the SCs. Theseprovisions largely have focused on affirmative action/ protective discrimina-tion for both the SCs and STs. Accordingly, a certain proportion of seatsin parliament/state legislatures, higher educational institutes and jobs inthe public sector are reserved for individuals belonging to the SC/ST socialgroups.12Untouchability was abolished and its practice was made a punishable offence by Article17 of the Indian Constitution (1949). Thereafter, Protection of Civil Rights Act, 1955laid down punishments for committing untouchability against the SCs. The ScheduledCaste and Scheduled Tribes (Prevention of Atrocities) Act, 1989 was passed by the IndianParliament as it found the provisions for punishments under the Protection of Civil RightsAct, 1955 inadequate to deal with more violent forms of caste-based offences.513.1. IntroductionThe previous literature has largely studied the SCs and STs as one sin-gle disadvantaged group and have compared how SC/STs together havefared relative to non-SC/STs 13. But this can potentially mask importantdifferences between the SCs and STs. For instance, improvement in the well-being of SC/STs could be driven by SCs alone. Sarkar et.al (2006) find thatthe Human Development Index (HDI) for STs is around 30% lower thanall-India average and is comparable to those for countries in sub-SaharanAfrica. However with the exception of largely descriptive studies or studiesconducted in other fields such as sociology 14, STs have not been studied inisolation from the SCs. Therefore, we study STs in isolation from the SCsand compare STs with their closest socio-economically deprived counterpart,the SCs.Significant debates exist among anthropologists and sociologists regard-ing whether STs can be described as indigenous people of India 15 althoughit is generally agreed upon that they differ from the SCs. Mitra (2008) men-tions that the STs can trace their origin to four distinct racial groups: theNegritos, the Proto-Austroloids, the Mongoloids and the Caucasoid. Thetribes are said to have inhabited India before the arrival of the Aryans whofound tribal culture primitive, thereby forcing the tribes to move to isolatedforested and mountainous areas. The tribes have remained historically iso-lated and therefore have socio-cultural values different from the mainstreamHindu society. On the other hand, Beteille (1986), Xaxa(1999) argue thatunlike the Americas and even Australia and New Zealand, identifying indige-nous people in India is complicated due to migration patterns of differentraces within the country at different points of time in history. Althoughmost tribes settled in India before the arrival of the Aryans 16, they cite theexamples of the settlement of some of the tribes in North-East India like theNagas (around the middle of first millennium BC), the Mizos (around 16th13Kijima (2006), Hnatkovska et. al. (2012, 2013) are examples of studies in economics.14Das, Kapoor and Nikitin, (2010); Maharatna (2000, 2004, 2011) are some examples.15The Scheduled Tribes are also called adivasis, which literally translates into “originalinhabitants” and is thought to be a more appropriate term for them relative to “indige-nous” (Beteille (1986)).16The tribes originating from the Negritos is one such example.523.1. Introductioncentury AD) and the Kukis (after the Mizo settlement), which took placeafter the arrival of the Aryans. Xaxa (1999) notes that different groupsdeveloped distinct association with certain territories during the course oftheir historical movements and illustrates that non-tribal groups like Ben-galis have developed strong attachment to Bengal, making them the originalsettlers of that territory. Thus rather than arguing that all tribes necessarilysettled before the arrival of the Aryans, it is more important to recognizethat they have preferential rights over the territory which they inhabited dueto their historical settlement patterns (Xaxa (1999)). Unfortunately theserights of tribal groups have not been recognized by other dominant commu-nities in India, thereby resulting in the need for tribal groups to claim thatthey are the indigenous people of the country (Xaxa (1999)).Guha (2007) mentions the Dhebar Commission Report (1960-1961) whichidentified the continuation of colonial forest laws aimed at suppressing tribalrights over forests, even after independence, a leading cause of the loss of con-trol of tribals over forested lands. Further, the Commission also mentionedthat development projects such as power and steel plants during the SecondFive Year Plan (1956-1961) had resulted in large scale displacement of tribalsfrom their land. Other forest laws of independent India that aimed at for-est conservation, thereby restricting access to forests, had adversely affectedtribes whose traditional livelihood depends on forest use (Baviskar (1994)).Displacement and the consequent plight of the tribals due to building ofpower plants, dams, mining and industries have been widely documented(Baviskar (1995; 1997), Ekka (2000), Ekka and Asif (2000), Maharatna andChikte (2004), Sundar (2009)). On the other hand the SCs have historicallybeen an integral part of the mainstream Hindu society, although they havebeen oppressed by the upper caste Hindus. Although denied access, SCshave been exposed to information, technology, employment opportunitiesas they have lived in close proximity to the dominant Hindu communities(Xaxa (2001)). Whereas the disadvantage faced by the SCs arises from theirlow social status in the mainstream Hindu society, the disadvantage facedby the STs have mostly arisen from their isolation and subsequent loss ofcontrol over territory and other natural resources. Thus, the situation faced533.1. Introductionby the STs is quite distinct from that faced by the SCs. This provides moti-vation to study STs and SCs separately. It is in this regard that this chapterattempts to make a contribution to the existing literature in economics thathas often grouped STs and SCs as a single disadvantaged group in India.The existing literature has largely focused on education and labour mar-ket outcomes of various social groups. Kijima (2006), Hnatkovska et. al.(2012, 2013) have studied educational and occupational mobility betweenthe SC/STs and non-SC/STs over time as well as across generations. How-ever, the health situation of the STs vis-a-vis the SCs have remained rela-tively understudied 17. Therefore, this study aims at making a contributionto the existing literature in economics by studying health, which is not onlya key indicator of development but also a major public good that the stateprovides. We explore the differences not only between the SC/STs and non-SC/STs, but specifically between the STs and SCs themselves along withthe plausible reasons for differences in such health outcomes.In this chapter, we study differences between STs and SCs in terms ofwomen’s contraceptive knowledge and use, utilization of pre and postnatalhealthcare, awareness about diseases like HIV/AIDS and tuberculosis (TB),nutritional status as well as children’s health at birth, mortality and immu-nization status. We find that STs perform poorly relative to SCs in nearlyall the aforementioned health outcomes. We, then, investigate the plausi-ble mechanisms that can explain this health gap. In particular, we studywhether it is the lack of demand for healthcare from the STs or the dearthof supply in public health services that can explain these health differences.We consider education, exposure to the media, access to basic householdamenities (like drinking water, electricity and flush toilet) and women’s so-cial status as potential factors that can influence the demand for healthcare.17Descriptive analysis on tribal health are found predominantly in the literatures ofsociology, anthropology and medicine. Ramaiah (2015) provides a descriptive analysis ofthe nutritional status of and access to healthcare for SCs and STs. Maharatna (2011)compares the crude death and infant mortality rate between the STs and SCs amongother outcomes such as per capita consumption, landownership and literacy. Singh et. al.(2014) study the nutritional status of women belonging to the Bhaina Tribe of CentralIndia. Das and Bose (2015) study nutritional status for tribes in terms of the prevalenceof chronic energy deficiency and low body mass index.543.2. DataAvailability of public health infrastructure and medical personnel are con-sidered as indicators of supply of health services. We find that althoughthe STs are somewhat less educated and exposed to the media relative tothe SCs, ST women enjoy greater autonomy than SC women in general.Hence, the conjecture that poor social status of women can contribute tothe poor state of women’s and children’s health is unlikely to hold for theSTs. Further, using Blinder-Oaxaca decomposition technique we find thatit is largely the supply of health services that seems to play a relativelyimportant role rather than demand for healthcare in explaining the healthdisparity between the STs and SCs. We also find that the National RuralHealth Mission (NRHM) implemented by the Government of India since2005 is unlikely to be able to reduce the scarcity in health infrastructure inrural, tribal areas. Therefore, a major policy implication of this study is notonly to strengthen the demand for healthcare, but focus more on improvingthe supply of health infrastructure in tribal areas.This chapter is organized as follows: Section 3.2 describes the data usedand the descriptive statistics; Section 3.3 presents the empirical specifica-tions used; Section 3.4 presents the results; Section 3.5 describes the plausi-ble reasons that might explain the health gap; Section 3.6 briefly discussesthe possible role of the National Rural Health Mission (NRHM) in bridgingthe health disparity between the STs and non-STs and Section 3.7 concludes.3.2 DataThe data used for analysis in this chapter come from two rounds of theDemographic and Health Surveys (DHS) conducted in 1998-99 and 2005-06.The data from 2005-06 are used to compare whether our outcome variablesof interest have changed over time for the different social groups. This ismotivated by the fact that the Government of India formulated a NationalHealth Policy in 2002 (the preceding National Health Policy was formulatedin 1983). It prescribed that state governments should increase the share oftheir expenditures to health from 5.5% to 7% by 2005. It also aimed ateradicating polio and leprosy, increase government expenditures for medical553.2. Dataresearch that would aid in developing therapeutic drugs as well as vaccinesagainst malaria, tuberculosis and HIV/AIDS by 2005 (Ministry of Healthand Family Welfare, Government of India). The data on health outcomesof women have been collected from ever-married women in the ages of 15-49 years. During the Round 2 of the DHS (1998-99), about 12% of thewomen respondents were Scheduled Tribes (ST), 17% were Scheduled Castes(SC), 29% were Other Backward Classes (OBC) and the remaining belongedto non-SC/ST/OBCs (which includes the upper castes). The Round 3 ofthe DHS (2005-06) comprised of 14 % ST, 17% SC, 33% OBC women andthe remaining women belonging to non-SC/ST/OBC category. We providedescriptive statistics of key demographic variables in Table 3.1. The DHSrecords detailed demographic information such as the woman’s age in thesurvey year, her year of birth, whether she resides in rural or urban area,the district (only in 1998-99) and the state (both in 1998-99 and 2005-06)in which she currently lives and her religion. We find that the mean ageof the respondent at the time of the survey was around 30 years in 1998-99 and it fell slightly to 28 years in 2005-06. Although individuals aremore likely to live in rural areas and the proportion of individuals living inrural areas declined somewhat from 1998-99 to 2005-06, we find that on anaverage 71% ST women lived in rural areas compared to 56% of SC womenin 2005-06. Both ST and SC women are more likely to report themselves asHindus rather than Muslims; however, ST women are more likely to reportthemselves as Christians relative to SC women.3.2.1 Health OutcomesThe outcomes that we consider are the respondent’s knowledge and usageof modern contraception; awareness about HIV and TB; usage of iodizedsalt in food; whether the respondent’s youngest child was born “small”; therespondent is anaemic, her last child’s immunization and children’s mortal-ity. We consider outcomes such as whether the child was born “small” andimmunization status for the last child born during the three years preced-ing the survey year. This is because the DHS provides detailed information563.2. Dataon these outcomes only for the last and second-last born child for Round2 and for the last, second-last and third-last born child for Round 3 of thesurvey who were born during the three years before the respective surveyyears; and not for all of the respondent’s children. Now, the sample sizepertaining to these outcomes for the last-born child is larger than that forthe second or third-last born child. Therefore, we use the last-born childonly for our analysis regarding these outcomes. We use Round 2 of the DHSto obtain information on the complete birth history of each respondent inorder to construct our child mortality indicators because this round pro-vides district identifiers unlike Round 3, where this information has beensuppressed for confidentiality purposes. The complete birth history recordsthe year of birth of each child ever born to the woman, the childs gender,the birth order, whether the child is one of a multiple birth, whether thechild has died and the age of death. This information is used to computemortality outcomes for children like the prevalence of neonatal, infant andchild mortality. As the data on children born in the 1960s are rather thin,we restrict the data to include only those children who were born between1970 and 1999.Table 3.1 reports the summary statistics of some of these variables whichare available from both rounds of the DHS. We find that knowledge of mod-ern contraception is high overall; although SCs are more likely to knowabout contraceptives relative to STs. Likewise, the usage of contraceptivesis higher among SC than ST women. We find a somewhat decline in knowl-edge and usage of contraceptives among both SCs and STs between the twoDHS rounds. On an average, 28% ST children were born “small” relative to24% among the SCs in 1998-99 and this appears to have declined over timebetween the rounds. Also, ST children are around 10% less likely to receiveBCG, measles, all doses of DPT and polio vaccines relative to SC childrenin both rounds of the DHS. Although immunization coverage appears tohave increased over time, the gap between ST and SC children in terms ofimmunization seems to be high. Figures 3.1, 3.5 and 3.6 depict trends inknowledge and usage of modern contraception and children’s immunizationstatus for different social groups across the two DHS rounds. Also on an573.2. Dataaverage, 4% ST and SC children have died within 1 month of birth (neonatalmortality), 8% ST and SC children have died within 1 year of birth (infantmortality), 11% ST children and 12% SC children have died within 5 yearsof birth (child mortality).3.2.2 Utilization of Health ServicesHere, we consider whether the respondent was given tetanus toxoid injectionand folic tablets during her last pregnancy. Also, we consider whether shegave birth to her last child at home or in a government medical institution.Table 3.1 shows that around 61% ST women received a tetanus toxoidinjection during their last pregnancy compared to 75% SC women in 1998-99. This increased to 72% for ST women and 84% for SC women in 2005-06;thereby indicating that the gap between ST and SC women in receiving thisinjection still continues to be large. Further, we also find that whereas thefraction of ST women giving birth at home with no medical assistance atdelivery went down from 78% to 72% between 1998-99 and 2005-06; theproportion of SC women giving birth at home reduced from 73% in 1998-99to 59% in 2005-06. Figures 3.2 and 3.4 depict trends in antenatal care andplace of delivery for different social groups across DHS rounds.3.2.3 MechanismsIn order to study the potential mechanisms, we consider differences in ed-ucation; exposure to the media in the form of watching TV and listeningto radio; access to basic amenities such as drinking water, flush toilet andelectricity and women’s social status as indicated through labour marketparticipation, ability to make intra-household decisions and freedom of mo-bility, across women belonging to different social groups. These factors arelikely to influence the demand for health services. For the purpose of un-derstanding the role of supply of health infrastructure, we consider whetherthe respondent received antenatal care and assistance at the time of de-livery from professional healthcare providers, the quality of prenatal carereceived (such as whether her weight was measured, blood pressure checked583.3. Empirical Specificationand blood test done) as well as experiences obtained when seeking care ata health facility/health camp (such as availability of health service sought,cleanliness of the facility and behaviour of the medical staff).Figure 3.1 shows trends in visits by family planning worker to respon-dents’ homes, Figures 3.2 and 3.3 depict the quality of antenatal care andcare received at the time of giving birth for women belonging to differentsocial groups across DHS rounds.3.2.4 Health Infrastructure under NRHMIn order to study trends in the availability of medical infrastructure in ruraltribal areas under the National Rural Health Mission (NRHM) implementedsince 2005, we use data on the shortage in the availability of medical facili-ties (computed by taking into consideration tribal population in rural areasaccording to the Census, 2001) for the period 2005-2010 from the RuralHealth Statistics published by the Ministry of Health and Family Welfare,Government of India. We use information on the dearth in the availability ofhealth sub-centres, primary health centres, community health centres, doc-tors, specialists (obstetricians, gynaecologists, surgeons and paediatricians),pharmacists, lab technicians and trained nurses/midwives as possible indi-cators of health infrastructure in rural, tribal areas for our analysis.3.3 Empirical SpecificationIn this section, we study the differences in health outcomes not only betweenSTs and non-ST/SC/OBCs which also includes the upper castes; but alsobetween the STs and SCs themselves. The following estimation equationsare used:yisr = α+ β0STisr + β1SCisr + β2OBCisr + γXisr + δs + εisr (3.1)Here yisr refers to the outcome variable of woman (or child) i in state s inthe DHS round r, where round refers to Round 2 (1998-99) or Round 3 (2005-593.3. Empirical Specification06). This estimation equation is used when data on yisr are available onlyfor one of the two rounds. STisr,SCisr and OBCisr are dummy variablesthat take the value 1 if the respondent or her child is ST, SC or OBCrespectively and 0 otherwise. Respondents belonging to non-ST/SC/OBCsocial group that also includes upper castes is the omitted category. Xisr isthe vector of control variables that includes dummies for religion, such as ifthe individual is Hindu, Muslim, Christian or Sikh; a dummy for whetherthe respondent lives in rural area. δs refers to state fixed effects and εisr isregression disturbance term that is clustered at the district level for Round2 and at the state level for Round 3 (as Round 3 does not report districtlevel identifiers).yisr = α+β0STisr+β1SCisr+β2OBCisr+γXisr+κR3+θsocSocGroupisr∗R3+δs+εisr(3.2)This equation is used to estimate whether there have been any changes inthe outcome variables of interest across rounds as well as between differentsocial groups across rounds. Pooled data from two rounds of the DHS areused for this purpose. Here, R3 assumes the value 1 if the respondent wasinterviewed during Round 3 (2005-06) and is 0 otherwise. SocGroupisr ∗R3is the interaction of the dummy for round and the dummy for the respon-dent or her child’s social group (if ST, SC or OBC and non-ST/SC/OBCis the omitted category for social group). Regression disturbance term εisris clustered at the state level as Round 3 data are also used for estimationhere.yisr = α+β0STisr+β1SCisr+β2OBCisr+γXisr+λBCisr+ηsocSocGroupisr∗BCisr+δs+εisr(3.3)For some outcome variables, we also estimate whether the outcome vari-ables have changed across different birth cohorts of women and betweendifferent social groups across birth cohorts. The estimation equation (3)is used for this purpose. Here, BCisr refer to birth cohort dummies for603.4. Resultswoman i in state s in the DHS round r, based on the individual’s year ofbirth. SocGroupisr ∗ BCisr refer to interaction between the respondent’ssocial group and her birth cohort dummies.3.4 ResultsHere we describe the results. At first, we present the results pertaining tomodern contraception, followed by those on antenatal care, delivery care,awareness about HIV and TB, usage of iodized salt in food, child’s health atbirth, women’s nutritional status, children’s mortality and lastly children’simmunization.3.4.1 Modern ContraceptionModern contraceptive methods aid in family planning and provide protectionfrom sexually transmitted infections to some extent. Modern contraceptivemethods include oral contraceptive pills, condoms, IUD, male and femalesterilization. Table 3.2 depicts the differences between women belongingto different social groups in terms of their knowledge and usage of moderncontraceptive methods. Columns (1) and (2) use data from DHS Round 2(1998-99) where Column (2) includes state fixed effects as well. Columns(3) and (4) use pooled DHS data from Rounds 2 and 3 (1998-99; 2005-06),where Column (4) also estimates whether trends in modern contraceptiveknowledge and usage have changed between the two DHS rounds as well asbetween different social groups across the rounds. Thus, estimation equation(1) is used for estimation in Columns (1) and (2) and estimation equation(2) is used for estimation in Column (4). We focus on Columns (2) and(4) in Table 3.2 for the purpose of interpretation of the results. Here, theomitted category for social group is non-ST/SC/OBC women.From Column (2) in the upper panel of Table 3.2, we find that STsare 1.8% less likely to know about modern contraceptives relative to non-ST/SC/OBC women, whereas there appears to be no significant differencebetween SC and non-ST/SC/OBC women in this regard. This implies that613.4. ResultsST women are 1.8% less likely to know about modern contraceptive methodsrelative to SC women as well. Column (4) shows that although ST womenare around 1.7% less likely to know about modern contraceptives, thereappears to be no difference in terms of this outcome variable across theDHS rounds or between women of various social groups across rounds.Column (2) in the lower panel of Table 3.2 shows that ST women are11.5% less likely to use modern contraceptives in contrast to SC women whoare 7.9% less likely to use modern contraceptives relative to non-ST/SC/OBCwomen. Therefore, ST women are around 3.5% less likely to use moderncontraceptives relative to SC women. As in the upper panel of Table 3.2,Column (4) in the lower panel of Table 3.2 shows that although ST womenare around 3% less likely to use modern contraceptives compared to SCwomen, there appears to be no difference in terms of this outcome variableacross the DHS rounds or between women of various social groups acrossrounds.Therefore, ST women are not only less likely to know or use modern con-traception methods relative to non-ST/SC/OBC women which also includesupper castes, but also relative to SC women 18.3.4.2 Antenatal Care, Infant Health, AwarenessWe focus on the probability of obtaining tetanus toxoid injection and folictablets during the respondent’s last pregnancy as outcome variables for ante-natal care. We consider the place of delivery and whether the child was born“small” as indicators of care obtained at delivery and infant health. Givingbirth at home is largely associated with receiving no professional medicalassistance. As government institutions would largely account for most ofthe institutional deliveries, we focus on the probability of giving birth atgovernment medical institutions in this analysis. The DHS mentions thatmost children are not weighed at birth in India. Therefore, mothers wereasked whether they thought their child was born “large, average, small or18These results are robust to also controlling for the respondent’s years of education.We do not usually control for years of education in these baseline specifications becauseeducational attainment itself could be endogenous to an individual’s caste/tribe.623.4. Resultsvery small”. The DHS reports also mention that children who are perceivedby their mother to be born “small” face increased risk of mortality. As such,the quality of antenatal care received can potentially influence whether thechild was born “small”. We also include variables that indicate whetherwomen use iodized salt in food and are aware of HIV, TB. The World HealthOrganization (WHO) recommends getting immunized against maternal andneonatal tetanus. Also WHO recommends daily iron and folic acid supple-mentation during pregnancy in order to avoid iron deficiency, risk of lowbirth weight and maternal anaemia. Therefore, tetanus injection and folicacid consumption constitute important aspects of antenatal care. Althoughconsumption of iodine and awareness about HIV and TB are important forone’s health in general, iodine supplementation during pregnancy is recom-mended by WHO for proper foetal development. Also, mother’s knowledgeof HIV and TB are important as mother-to-child transmission of HIV canbe substantially reduced through appropriate medical intervention (WHO)and active TB infection during pregnancy needs to be treated because it canimpede foetal growth.The data on whether the respondent received tetanus injection, place ofdelivery and whether the child was born “small” are available for both therounds. The information on whether she got folic tablets is only availablefrom Round 2 as the survey question was altered between the two rounds,making the information non-comparable across rounds. Data on the use ofiodized salt and awareness about TB are available only for Round 3 and onknowledge of HIV from Round 2. The results on these outcome variableshave been reported in Table 3.3. For the upper panel in Table 3.3, we usethe regression specification as in Column (4) of Table 3.2. On the otherhand for the lower panel, we use the regression specification of Column (2)of Table 3.2.From the upper panel of Table 3.3 we find that ST women are around13% less likely to have obtained tetanus injection during their last pregnancyrelative to SC women. ST women are around 6% more and 6% less likely togive birth at home and government medical institution respectively relativeto SC women. We also find that ST children are 2% more likely to be born633.4. Resultssmall relative to their SC counterparts. Further, trends in these outcomevariables do not seem to vary significantly over rounds or over different socialgroups across the DHS rounds.The lower panel of Table 3.3 shows that ST women are 7% less likely tohave consumed folic tablets during their last pregnancy and use iodized saltin food. They are also 9% and 13% less likely to be aware of HIV and TBrespectively relative to SC women.3.4.3 Women’s Nutritional StatusWe study the differences between ST and SC women in the probability ofbeing anaemic in the upper panel of Table 3.4.We use estimation equation (3) in estimating the prevalence of anaemiaacross birth cohorts of women in Column (3) of Table 3.4. Column (3) ofTable 3.4 shows that ST women are around 7% more likely to be anaemicthan SC women and the prevalence of anaemia does not appear to be sig-nificantly different across birth cohorts of different social groups. Anotherindicator of nutritional status, namely being underweight has been consid-ered in Appendix B. Column (2) of Appendix Table B.1 shows that STwomen are 1.5% more likely to be underweight relative to SC women. Ingeneral, younger women are less likely to be underweight and younger STwomen are slightly less likely to be underweight than their SC counterparts.3.4.4 Children’s MortalityWe consider the probability of a child dying within 1 month (neonatal mor-tality), 1 year (infant mortality) and 5 years (child mortality)of birth byconsidering the entire birth history of the respondents from Round 2 of theDHS. We restrict our data to include only those births at the respondent’scurrent place of residence to mitigate the possibility of selective migrationto other districts/states that might provide better healthcare for newborns.We include the child’s year of birth fixed effects to control for overall changesover time in healthcare for children and state fixed effects to take into ac-count time-invariant differences across states in child health. Further, con-643.4. Resultstrols for whether the child was female, if she/he was one of a multiple birth,her/his birth order, mother’s age at birth, religion and rural dummies areincluded. We report the results on children’s mortality in the lower panel ofTable 3.4. Although we find that neonatal, infant and child mortality arehigher among the ST/SC relative to their non-ST/SC/OBC counterparts,ST children are less likely to face infant and child mortality in particularrelative to SC children. This finding appears to be supported by Maharatna(2000) who quotes from other studies that living in less environmentallypolluted areas (such as forests), the customs of prolonged period of breast-feeding and greater birth spacing among tribes appear to explain why infantand child mortality are lower among the STs than the SCs.3.4.5 Children’s Immunization StatusChildren’s immunization against diseases such as BCG, measles, DPT andpolio are considered to be important aspects of child health. Table 3.5 showsthat ST children are around 7% less likely to have obtained BCG or measlesvaccines relative to SC children. They are also around 9% less likely to havereceived all three doses of the DPT vaccine and 3% less likely to have receivedall three doses of the polio vaccine, compared to SC children. Particularlythe probability of receiving all doses of polio vaccine seems to have improvedacross DHS rounds and this could be attributable to the extensive pulse-polio vaccination campaigns run by the government. However, ST childrenare around 17% less likely to have received all doses of polio vaccine in2005-06; whereas SC children are only 9% less likely to have received allpolio doses in 2005-06 relative to non-ST/SC/OBC children. Overall, STchildren are 8% less likely to have received any vaccine at all and around10% less likely to have a health card that keeps a record of the child’simmunization history. Thus, there still appears to be large gaps in childimmunization status between ST and SC children and this does not seem tohave declined over time.653.5. Plausible Mechanisms3.5 Plausible MechanismsThere might be several plausible reasons that might explain the differences inhealth outcomes of women and children belonging to the Scheduled Castesand Scheduled Tribes. For example, education and exposure to the me-dia can influence one’s awareness about and hence demand for health andmedical care. Access to basic household amenities such as drinking water,toilet at home and electricity can also influence hygiene as well as knowl-edge about health. Therefore, differences in education, exposure to mediaand access to basic household amenities can potentially affect differences inhealth outcomes between STs and SCs. As the current analysis focuses onhealthcare of women and children, it is also important to analyse the statusof women in society. This is because low social status of women can influ-ence women’s ability to seek healthcare and this can in turn influence thehealth outcomes of women and children. Lastly, the role of medical facilitiesthemselves need to be analysed to understand whether lack of availabilityor supply of health infrastructure can contribute largely to the disparity inhealth outcomes between STs and SCs.3.5.1 Differences in EducationTable 3.6 depicts the differences in the probability of being illiterate andyears of education among women. We find that ST women are somewhatmore likely to be illiterate than SC women. However, whereas the youngestcohort of ST women are 6% less likely to be illiterate; the youngest cohortof SC women are around 13% less likely to be illiterate. We also find thatthe years of education is lower by around 0.5 years for the youngest cohortof ST women; while there appears to be no significant difference in years ofeducation between the youngest cohort of SC and non-ST/SC/OBC women.Therefore, although differences in education appear to be small, it can stillsomewhat influence the poor health status of ST women and children vis-a-vis their SC counterparts.663.5. Plausible Mechanisms3.5.2 Differences in Household AmenitiesThe upper panel of Table 3.7 reports the results on the differences in accessto basic household amenities across social groups. We find that althoughboth SC and ST households lag behind in the availability of drinking water,flush toilet and electricity in their households relative to non-ST/SC/OBChouseholds, ST households are more likely to lack these basic amenitieseven relative to SC households. The upper panel of Table 3.7 shows thatST households are around 9%, 3%, 8% less likely to have drinking water,flush toilet and electricity in comparison to SC households. The availabilityof flush toilet in homes appears to have increased over time. Though SThouseholds appeared to have gained relative to SC households in havingdrinking water in the their premises in 2005-06, the same cannot be saidabout the availability of flush toilet and electrification of ST homes.3.5.3 Differences in Exposure to MediaFrom the lower panel of Table 3.7 we find that ST women are 26% less likelyto watch TV whereas SC women are 17% less likely to watch TV relative tonon-ST/SC/OBC women. Similarly, ST women are 15% less likely to listento radio while SC women are 12% less likely to listen to radio relative tonon-ST/SC/OBC women. Therefore, ST women are around 9% less likelyto watch TV and 3% less likely to listen to radio compared to SC women.ST women have, therefore, lower exposure to media compared to SC women.3.5.4 Differences in Women’s StatusWe consider women’s labour market participation, ability to solely makeintra-household decisions and freedom of mobility as potential indicators ofwomen’s social status.Table 3.8 shows that both ST and SC women are more likely to partici-pate in the labour market than non-ST/SC/OBC women that also includeswomen belonging to the upper castes. Further, ST women are 6% morelikely to be currently working and 5% more likely to be working outsidetheir homes relative to SC women. This echoes the findings in Eswaran et.673.5. Plausible Mechanismsal. (2013) where they find that married women belonging to upper castesare less likely to participate in the labour market relative to those belongingto the lower caste on account of concerns about family status that is moreprevalent among upper castes than among the lower castes. However, herewe find that ST women are by far most likely to work, even relative to lowercaste women. ST women are also more likely to work outside their homeseven relative to their SC counterparts.We also report the findings on variables that could plausibly indicatewomen’s autonomy in Table 3.8. We find that ST and SC women are equallylikely to decide how they are going to spend the wages they earned fromwork. However, ST women are not only more likely to solely decide whatto cook, to obtain healthcare for themselves and purchase jewellery relativeto non-ST/SC/OBC women (including upper caste women), but this is alsofound to hold relative to their SC counterparts. Further, relative to SC andwomen belonging to other social groups, ST women are more likely to goto the market and visit friends or relatives without seeking permission. Ap-pendix Table B.2 shows that infant and child mortality are higher amongmale rather than female ST children 19. Appendix Table B.3 shows thatST women are also more likely to have a greater proportion of daughters,daughters to sons and daughters to sons alive relative to women of all othersocial groups, including the SCs. These findings, therefore, indicate that STwomen enjoy a higher degree of autonomy in intra-household decision mak-ing as well as enjoy greater mobility relative to all other women, includingSC women 20.These findings suggest that tribal societies are less likely to practicegender discrimination against women. As gender discrimination againstwomen usually results in poor health outcomes of women and their inabilityto access healthcare; lower social status of women is unlikely to explain whyST women and children perform poorly in terms of health relative to their19Maharatna (2000) mentions that this phenomenon is observed among most tribes inIndia.20This is supported by Mitra (2008) who mentions that tribal societies generally havemore equal gender relations and thus have a relatively higher social status for women thannon-tribals in India.683.5. Plausible MechanismsSC counterparts.3.5.5 Differences in Medical CareHere we study whether differences in medical facilities and care can explainthe differences in health outcomes between ST and SC women and children.We consider differences in experiences about health facilities and person-nel, antenatal care received during the last pregnancy as well as medicalassistance received while giving birth in explaining the disparities in healthoutcomes between STs and SCs. Although the DHS records complete birthhistory of each respondent, it only records the quality of pre and postna-tal medical care received for the last two or three births within the threeyears preceding the survey year. However, as a relatively large number ofobservations are available for the last birth of the respondent than for thesecond-to-last births and so on, we use the quality of medical care receivedduring the last pregnancy and during the birth of the last child as outcomevariables of interest.Table 3.9 reports differences in experiences at health facilities that in-dividuals faced when they visited health facilities or camps for themselvesor their children during the last 12 months. We find that ST women areclose to 10% less likely to receive the health service that they sought at thehealth facility/camp, feel that the staff at the facility spent “enough” timewith them and find that the facility was clean. On the other hand, thereappears to be no significant difference between the experiences of SC womenand non-SC/ST/OBC women when they visited a health facility/camp.Table 3.10 shows that both ST and SC women are less likely to havereceived antenatal care from a doctor, from any professional health worker(doctor, trained nurse, other health workers etc.) or have an antenatal healthworker visit them in their homes relative to non-ST/SC/OBC women. Theyare also likely to receive care later in their pregnancy. However, there isdisparity between ST and SC women as well. For example, ST women are7% less likely to have received antenatal care from a doctor, 6% less likelyto have received healthcare from any professional healthcare provider, 6%693.5. Plausible Mechanismsmore likely to receive the care later in pregnancy and 7% less likely to havean antenatal health worker visit them in their homes relative to SC women.Table 3.10 also reports disparity in the quality of antenatal care received bywomen during their last pregnancy. We find that although both ST and SCwomen are less likely to have their weight measured, blood pressure checkedor have their blood test done during their last pregnancy; ST women are7%, 4% and 3% less likely to have their weight measured, blood pressurechecked and have their blood test done relative to SC women respectively.Further, trends in most of these outcomes do not appear to have improvedover time or for different social groups across DHS rounds.In Table 3.11 we find that ST women are 6% less likely to receive assis-tance from a doctor and 10% less likely to receive any professional medicalassistance (from a doctor, trained nurse or midwife) while giving birth totheir last-born children in comparison to SC women. They are also 8% morelikely to have a traditional (untrained) birth attendant and 6% more likelyto have a relative/friend assist them while giving birth in comparison to SCwomen. No improvement in these outcomes could be found over time oracross different social groups over time.3.5.6 The Role of Various FactorsIn this subsection, we analyse the relative importance of different factorsin explaining the differences in health outcomes between the STs and SCs.The previous subsections have shown that disparities in education, exposureto media, access to basic household amenities and medical care could plau-sibly explain why the STs have poorer health outcomes relative to SCs. Onthe other hand, the status of women in society is unlikely to explain thesedifferences as ST women appear to enjoy greater autonomy and mobilityrelative to even SC women. Therefore, given these findings we use Blinder-Oaxaca decomposition methodology to analyse the relative importance ofeducation, exposure to media, access to basic household amenities and med-ical care in explaining the health disparities between STs and SCs. We alsocontrol for the respondent’s age, whether she lives in rural area, her religion703.5. Plausible Mechanismsand state fixed effects in the linear regressions used for the Blinder-Oaxacadecomposition method.Table 3.12 reports the role of different factors in explaining the differencesin the knowledge of modern contraceptives between ST and SC women. Wefind that SC women are 5% more likely to know about modern contraceptivesrelative to ST women and this difference is almost completely explainedby attributes such as education, exposure to media, access to householdamenities and visits by family planning worker as well as age, whether therespondent lives in rural area, her religion and state fixed effects. Of theattributes, age, having TV and visits by family planning worker appear tobe important in explaining the difference between ST and SC women interms of their knowledge of modern contraception (in addition to state fixedeffects). Of the different attributes, visits by family planning worker aloneexplains around 8% of the difference in knowledge of modern contraceptionbetween SC and ST women.In Table 3.13, we study the differences in the usage of modern contra-ceptives between SC and ST women. We find that SC women are 12%more likely to use modern contraceptives relative to their ST counterparts.About 9% of the differences are explained by attributes. We find that ofthe differences explained by the attributes, around 33% can be explainedalone by visits by family planner health worker. Although visits by familyplanning health worker appears to be an important factor in explaining thedifference in knowledge and usage of modern contraceptives between SC andST women, we find that interaction with family planning worker is more im-portant in explaining the differences in contraceptive usage between ST andSC women.Table 3.14 shows the difference in getting a tetanus injection before thebirth of their youngest child between SC and ST women. We find thatSC women are 11% more likely to have received a tetanus injection as partof their antenatal care relative to ST women of which about 8% can beexplained by attributes alone. Of the different attributes that can potentiallyinfluence getting a tetanus injection before giving birth, getting professionalprenatal healthcare can explain about 75% of the explained difference.713.6. The Role of the NRHMTable 3.15 reports the difference in getting folic tablets for consump-tion before the birth of their youngest child between SC and ST women.We find that SC women are around 6% more likely to have received folictablets, which is entirely explained by attributes. We also find that althoughgetting professional prenatal care is likely to have a large contribution in ex-plaining the difference in folic tablet consumption (from the magnitude ofits coefficient), it is not however statistically significant.Although disparities in education, exposure to media (as measured byownership of TV, radio), access to basic amenities like drinking water, flushtoilet and electricity at home can potentially explain the differences in healthoutcomes between ST and SC women and children; the role of medical pro-fessionals or access to medical care appears to be relatively more importantin explaining the differences in some of the health outcomes between theSTs and SCs. In other words, relative to the demand for healthcare which isinfluenced by education, exposure to media, access to household amenities;the supply of health infrastructure appears to be relatively more importantin explaining the poor state of health of STs relative to the SCs in a numberof situations.3.6 The Role of the NRHMIn this section we briefly discuss the role of the National Rural Health Mis-sion (NRHM) in attempting to bridge the health gap between the tribalsand non-tribals. The NRHM was implemented by the Government of Indiain April, 2005 with a view to improving rural health infrastructure. Thiswas important as maternal and infant mortality rates are particularly highas well as the prevalence of anaemia, tuberculosis etc. Rural areas not onlylack medical facilities like primary health centres, but also face scarcity oftrained medical personnel such as doctors, specialists (obstetricians, gynae-cologists, paediatricians, surgeons), nurses, lab technicians and pharmacists.Therefore, the NRHM was an important policy that had the potential tobridge the rural-urban differences in the availability of public health infras-tructure. Further, as around 71% of STs in 2005-06 reside in rural areas723.6. The Role of the NRHM(from Table 3.1 and these figures are higher than the corresponding figuresfor SCs), the NRHM holds particular potential in improving the health out-comes of the tribal population. The Rural Health Statistics published bythe Ministry of Health and Family Welfare, Government of India reports theshortfall in the availability of health sub-centres, primary health centres andcommunity health centres which are basic medical institutions serving therural population. Further, the Rural Health Statistics also report the short-fall in the availability of doctors, specialists, lab technicians, trained nursesand pharmacists. Figure 3.7 plots the map of India by dividing the statesinto those with population of more than 20 million, which we term “largestates” and those with population of less than 20 million, termed “smallstates”. The map also shows states for which STs comprise at least 10%and those for which STs comprise less than 10% of their population. Figure3.7 shows that states predominantly in northern, central and eastern Indiaare typically large states with at least 10% population that is ST; whereasstates in North East India are largely small states with at least 10% of theirpopulation being STs. In fact, STs largely comprise more than 50% of thepopulation of a large number of North-East states.Figures 3.8-3.13 depict trends in the shortfall of medical infrastructureand trained medical personnel in rural tribal areas of large and small stateswith at least 10% population that is ST. The Rural Health Statistics com-putes the shortfall in health infrastructure by taking into consideration therural ST population of the respective states according to the last availablecensus figures (for the period 2005-2010, this corresponds to the Census fig-ures from 2001). We find that shortfall in the number of health sub-centres,primary health centres, community health centres, doctors, specialists (ob-stetricians, gynaecologists, paediatricians, surgeons), pharmacists, lab tech-nicians and trained nurses/midwives is higher in large than in small stateswith large ST population. Also, any decline in the shortfall of medical in-frastructure is observed for small states alone.Figures 3.14-3.19 plots the regional differences in the shortfall of medicalcentres and trained medical personnel in rural, tribal areas. We plot stateswith at least 10% population that is tribal. Therefore, we do not plot the733.7. Conclusionshortfall in health infrastructure in rural, tribal areas of South Indian statesas states in southern India typically have very low share of tribal population.We plot the shortfall in health infrastructure in rural, tribal areas of statesin Northern, Eastern, Western and North-Eastern India. The states con-sidered in Northern India are Jammu & Kashmir, Rajasthan and MadhyaPradesh; those in Eastern India are Chattisgarh, Jharkhand and Odisha; theone in Western India is Gujarat and those in North-East India are Sikkim,Assam, Arunachal Pradesh, Meghalaya, Manipur, Nagaland, Mizoram andTripura (Figure 3.7). States in North and East India are found to havelarge shortfall in public health services. It is to be noted that Western Indiahas a lower share of tribal population relative to states in North-EasternIndia. Therefore considering that North-Eastern states are predominantlytribal relative to those in Western India, the shortfall in medical centres andtrained health personnel is largely lowest in North-Eastern states among allstates with at least 20% population that is ST. Therefore, large regional im-balances continue to exist in rural health facilities in tribal areas even afterthe implementation of the NRHM. Thus, this potentially raises scepticismabout the efficacy of the NRHM to bridge the gap in health between tribalsand non-tribals at least in the near future.3.7 ConclusionIn this chapter, we have studied the differences in health outcomes not onlybetween STs and non-ST/SC/OBCs, which includes the upper castes; butalso specifically between the STs and SCs. In particular, we have focusedon women’s and children’s health. As the previous literature has largelycombined STs and SCs together as one large disadvantaged group and havecompared ST/SCs together with non-ST/SCs in terms of education, occupa-tion, consumption and wages; few rigorous empirical studies exist that haveanalysed the situation of the STs and SCs separately. Further, healthcareis one of the most important public goods provided by the state and hasremained largely unexplored in the past literature on tribal communities inIndia. This provides us the motivation to study health outcomes of STs743.7. Conclusionand SCs and the plausible reasons that could explain the disparity in healthoutcomes between these two social groups. We find that ST women andchildren appear to perform poorly in nearly all indicators of health relativeto their SC counterparts.Although STs and SCs have been historically disadvantaged and havebeen subjected to similar affirmative action policies such as reservations ofseats in parliament, legislatures, public sector jobs and educational insti-tutions; it is worth exploring why there exists large disparities in healthoutcomes between STs and SCs themselves. We find that STs are some-what more likely to be illiterate, less likely to be exposed to the media (thatis, lower ownership of TV, radio) and possess basic household amenitiessuch as electricity, drinking water and flush toilet. As education and ex-posure to the media can influence awareness about health and possessionof basic amenities at home can influence personal hygiene; it is possiblethat these aforementioned reasons can plausibly explain the poor health ofSTs relative to SCs. On the other hand, the poor social status of womencan prevent them from obtaining the healthcare they need and cause poorhealth outcomes for women and children. However, we find that ST womenhave greater autonomy in intra-household decision making, enjoy greatermobility outside their homes, participate in the labour market and femaleST children are not more likely to face greater mortality risk relative totheir male counterparts (neonatal, infant child mortality). Thus, poor so-cial status of women is unlikely to explain the health disparities between STand SC women. We also explore whether STs face lower quality of medicalcare relative to SCs. In particular, we find that ST women are less likelyto receive any professional antenatal care, receive care later in pregnancy,are more likely to have no antenatal health worker visit them, less likelyto have their weight measured, blood pressure checked and blood test doneduring pregnancy; relative to SC women. Also, ST women are less likelyto receive assistance from a professional health personnel during the timeof delivery. ST women are also less likely to obtain the health service thatthey sought, feel the medical staff spent adequate time with them and findthe facility clean when visiting a health facility. Using the Blinder-Oaxaca753.7. Conclusiondecomposition technique, we find that although lack of education, exposureto the media and possession of household amenities can somewhat explainthe poor health outcomes of STs relative to SCs, the availability of medicalfacilities is relatively more important in explaining the health gap betweenthese two social groups for a number of situations. Therefore, it appearsthat improving the state of public health infrastructure can potentially helpbridge the gap between the STs and SCs. However, the NRHM implementedby the Government of India since 2005 does not appear to have resulted inthe decline in the shortfall of medical facilities in rural, tribal areas of “large”states that are mostly concentrated in western, northern and eastern India.This potentially raises scepticism about the efficacy of the NRHM in bridg-ing the health disparities between the STs and non-STs. It is possible thatas STs reside in remote, forested and hilly areas, medical personnel may notbe keen on serving in these areas. One way to tackle this problem couldbe for the government to provide financial incentives to medical personnelin order to induce them to serve in remote, rural areas. This could poten-tially improve the functioning of health infrastructure under NRHM in areaslargely inhabited by the STs.763.8. Figures3.8 Figures0.2.4.6.81Proportion WomenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsModern ContraceptionKnows Contraception Uses ContraceptionHas Heard About HIV Family Planner Worker VisitedFigure 3.1: Contraception, DHS 1998-99 & 20050.2.4.6.81Proportion WomenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsAntenatal CareGot Tetanus Injection Got No Antenatal VisitCare from Doctor Got Care from ProfessionalFigure 3.2: Antenatal Care, DHS 1998-99 & 2005773.8. Figures0.2.4.6Proportion WomenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsPostnatal CareDoctor At Birth Professional Care at BirthTrad. Birth Attendant Relative/Friend at BirthFigure 3.3: Trends in Delivery Care, DHS 1998-99 & 20050.2.4.6.8Proportion WomenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsPlace of DeliveryAt Home At Govt Medical FacilityAt Pvt Medical FacilityFigure 3.4: Trends in Delivery Care, DHS 1998-99 & 2005783.8. Figures0.2.4.6.8Proportion ChildrenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsChild ImmunizationHas Health Card Got BCG VaccinneGot Measles VaccineFigure 3.5: Trends in Children’s Immunization, DHS 1998-99 & 20050.2.4.6.81Proportion ChildrenOBC SC ST Upper Castes1999 2005 1999 2005 1999 2005 1999 2005by Social GroupsChild ImmunizationGot All DPT Vaccines Got All Polio VaccinesGot Any VaccineFigure 3.6: Trends in Children’s Immunization, DHS 1998-99 & 2005793.8. Figureskkkkkkkkkkkkkkkkkkkkkkkkkkk¯State-Wise Tribal Populationk >20 Million Total Populationk <20 Million Total PopulationNot Classified<10% Population Tribal>10% Population TribalFigure 3.7: Tribal Population by States, Census 2011803.8. Figures0 200 400 600Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Large States with atleast 10% population TribalShortfall of Health Centres in Rural Tribal AreasHealth SubCentre Primary Health CentreCommunity Health CentreFigure 3.8: Shortfall in Medical Facilities in Tribal Areas0 20 40 60 80Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Small States with atleast 10% population TribalShortfall of Health Centres in Rural Tribal AreasHealth SubCentre Primary Health CentreCommunity Health CentreFigure 3.9: Shortfall in Medical Facilities in Tribal Areas813.8. Figures0 100 200 300Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Large States with atleast 10% population TribalShortfall of Doctors in Rural Tribal AreasDoctors SpecialistsFigure 3.10: Shortfall of Doctors/Specialists in Tribal Areas0 20 40 60 80Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Small States with atleast 10% population TribalShortfall of Doctors in Rural Tribal AreasDoctors SpecialistsFigure 3.11: Shortfall of Doctors/Specialists in Tribal Areas823.8. Figures0 200 400 600 800 1,000Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Large States with atleast 10% population TribalShortfall of Medical Personnel in Rural Tribal AreasPharmacist Lab TechniciansNurses/MidwivesFigure 3.12: Shortfall of Other Medical Personnel in Tribal Areas0 50 100Shortfall20102009200820072005Source: Rural Health Statistics, Govt. of India (2005-2010)Small States with atleast 10% population TribalShortfall of Medical Personnel in Rural Tribal AreasPharmacist Lab TechniciansNurses/MidwivesFigure 3.13: Shortfall of Other Medical Personnel in Tribal Areas833.8. FiguresENNE WSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Primary Health Centre in Rural Tribal AreasFigure 3.14: Regional Differences: Shortfall in Medical Facilities in TribalAreasENNEWSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Public Health Subcentres in Rural Tribal AreasFigure 3.15: Regional Differences: Shortfall in Medical Facilities in TribalAreas843.8. FiguresENNEWSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Doctors in Rural Tribal AreasFigure 3.16: Regional Differences: Shortfall of Doctors in Tribal AreasENNEWSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Specialists in Rural Tribal AreasFigure 3.17: Regional Differences: Shortfall of Specialists in Tribal Areas853.8. FiguresENNEWSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Pharmacist in Rural Tribal AreasFigure 3.18: Regional Differences: Shortfall of Pharmacists in Tribal AreasENNEWSource: Rural Health Statistics, Govt. of India, 2005-2014States with atleast 10% population TribalShortfall in Lab Technicians in Rural Tribal AreasFigure 3.19: Regional Differences: Shortfall of Lab Technicians in TribalAreas863.9. Tables3.9 TablesTable 3.1: Summary StatisticsScheduled Tribes 1998-99 2005-06Variable Mean SD Obs Mean SD ObsWoman’s Age 30.64 8.55 10906 28.49 9.34 16537Rural 0.83 0.38 10906 0.71 0.45 16537Hindu 0.60 0.49 10890 0.42 0.49 16537Muslim 0.01 0.09 10890 0.02 0.13 16537Christian 0.29 0.45 10890 0.46 0.50 16537Knows Contraception 0.95 0.21 10906 0.93 0.26 16537Uses Contraception 0.38 0.49 10906 0.33 0.47 16537Tetanus During Pregnancy 0.61 0.49 4176 0.72 0.45 5733Gave Birth at Home 0.78 0.41 4176 0.72 0.45 5733Child was Born Small 0.28 0.45 4177 0.21 0.41 5544Child Got BCG Vaccine 0.54 0.50 3946 0.67 0.47 5475Child Got Measles Vaccine 0.26 0.44 3866 0.40 0.49 5353Child Got All DPT Vaccines 0.32 0.47 3908 0.42 0.49 5459Child Got All Polio Vaccines 0.04 0.21 3957 0.20 0.40 5495Scheduled Castes 1998-99 2005-06Variable Mean SD Obs Mean SD ObsWoman’s Age 30.73 8.81 15256 28.81 9.47 20397Rural 0.74 0.44 15256 0.56 0.50 20397Hindu 0.91 0.29 15235 0.88 0.32 20397Muslim 0.02 0.12 15235 0.02 0.14 20397Christian 0.03 0.16 15235 0.02 0.15 20397Knows Contraception 0.99 0.11 15256 0.98 0.15 20397Uses Contraception 0.46 0.50 15256 0.45 0.50 20397Tetanus During Pregnancy 0.75 0.43 5274 0.84 0.36 6331Gave Birth at Home 0.73 0.44 5273 0.59 0.49 6331Child was Born Small 0.24 0.44 5271 0.23 0.23 6251Child Got BCG Vaccine 0.64 0.48 4963 0.77 0.42 6049Child Got Measles Vaccine 0.34 0.48 4850 0.51 0.50 5941Child Got All DPT Vaccines 0.42 0.49 4937 0.53 0.50 6025Child Got All Polio Vaccines 0.09 0.29 4990 0.37 0.48 6044Note: Data source is Demographic and Health Surveys or DHS (1998-99; 2005-06). “SD” refersto standard deviation; “Obs” refers to the number of observations.873.9. TablesTable 3.2: Woman Knows/Uses Modern ContraceptionKnows Modern Contraception (1) (2) (3) (4)Scheduled Tribe -0.023*** -0.018*** -0.023*** -0.017***(0.004) (0.004) (0.005) (0.004)Scheduled Caste -0.002 -0.002 -0.004*** -0.003(0.002) (0.002) (0.001) (0.002)OBC 0.001 0.001 -0.002 -0.001(0.001) (0.001) (0.001) (0.002)Rural -0.009*** -0.009*** -0.009*** -0.009***(0.001) (0.001) (0.002) (0.002)Round 2005 -0.007(0.005)Round 2005*ST -0.008(0.011)Round 2005*SC -0.001(0.003)Round 2005*OBC -0.001(0.004)Uses Modern Contraception (1) (2) (3) (4)Scheduled Tribe -0.122*** -0.115*** -0.091*** -0.102***(0.016) (0.012) (0.019) (0.018)Scheduled Caste -0.098*** -0.079*** -0.052*** -0.070***(0.010) (0.008) (0.012) (0.016)OBC -0.063*** -0.040*** -0.027*** -0.033*(0.010) (0.007) (0.008) (0.017)Rural -0.114*** -0.072*** -0.048** -0.049**(0.010) (0.010) (0.020) (0.020)Round 2005 -0.035(0.046)Round 2005*ST 0.023(0.031)Round 2005*SC 0.034(0.024)Round 2005*OBC 0.015(0.024)Religion Dummies Yes Yes Yes YesWoman’s Age Yes Yes Yes YesState Fixed Effects No Yes Yes YesObservations 89577 89577 208,222 208,222Note: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variabledescription is true and is 0 otherwise. Data source is DHS (1998-99) for columns (1), (2) and pooledDHS (1998-99; 2005-06) for columns (3), (4). Robust standard errors clustered at the district level arein parentheses in columns (1), (2) and clustered at the state level are in parentheses in columns (3), (4).***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significance respectively.Regressions weighted by survey weight. Omitted category is non-SC/ST/OBC social group. Religiondummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh; the omittedcategory being all other religions.883.9. TablesTable 3.3: Woman’s Awareness, Prenatal Care, Infant HealthPanel A: Tetanus Home Institutional Child BornInjection Birth Birth (Govt) “Small”Scheduled Tribe -0.199*** 0.190*** -0.069*** 0.044***(0.023) (0.027) (0.019) (0.016)Scheduled Caste -0.074*** 0.129*** -0.016 0.025**(0.017) (0.024) (0.021) (0.010)OBC -0.051*** 0.042 -0.003 0.026**(0.013) (0.037) (0.016) (0.010)Rural -0.112*** 0.331*** -0.119*** 0.025***(0.019) (0.021) (0.016) (0.008)Round 2005 0.049 -0.050 -0.024 -0.026(0.045) (0.074) (0.037) (0.022)Round 2005*ST 0.042 0.126** -0.047 -0.010(0.043) (0.048) (0.034) (0.023)Round 2005*SC 0.006 0.059* -0.005 -0.013(0.025) (0.035) (0.029) (0.014)Round 2005*OBC 0.029* 0.090* -0.045* -0.024(0.016) (0.051) (0.024) (0.014)Observations 63868 63864 63864 63346Panel B: Folic Use Iodized Knows KnowsTablets Salt in Food HIV TBScheduled Tribe -0.158*** -0.127*** -0.256*** -0.174***(0.018) (0.027) (0.015) (0.019)Scheduled Caste -0.083*** -0.053*** -0.168*** -0.049***(0.012) (0.013) (0.010) (0.008)OBC -0.057*** -0.042*** -0.122*** -0.038***(0.010) (0.013) (0.010) (0.012)Rural -0.139*** -0.175*** -0.333*** -0.104***(0.012) (0.021) (0.014) (0.014)Observations 28647 112,412 89577 118,631Religion Dummies Yes Yes Yes YesWoman’s Age Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variabledescription is true and is 0 otherwise. Data are on last pregnancy during three years preceding the survey,except for use of iodized salt, awareness about HIV and TB. Data source is pooled DHS (1998-99; 2005-06)for Panel A where robust standard errors clustered at the state level are in parentheses. Data for folic tabletconsumption and knowledge about HIV are from DHS (1998-99) and standard errors have been clustered at thedistrict level for these outcomes. Data for other outcomes in Panel B are from DHS (2005-06) and standarderrors have been clustered at the state level for these outcomes. ***, ** and * indicate statistical significanceat the 1% , 5% and 10% level of significance respectively. Regressions weighted by survey weight. Omittedcategory is non-SC/ST/OBC social group. Religion dummies included control for whether the respondent isHindu, Muslim, Christian or Sikh; the omitted category being all other religions.893.9. TablesTable 3.4: Woman is Anaemic, Children’s MortalityWoman is Anemic: (1) (2) (3)Scheduled Tribe 0.161*** 0.163*** 0.155***(0.026) (0.022) (0.039)Scheduled Caste 0.068*** 0.067*** 0.076***(0.010) (0.006) (0.019)OBC 0.023 0.024*** 0.040**(0.018) (0.007) (0.018)Birth Cohort 2*ST -0.005(0.036)Birth Cohort 3*ST 0.006(0.027)Birth Cohort 4*ST 0.018(0.030)Birth Cohort 2*SC -0.022(0.020)Birth Cohort 3*SC -0.012(0.018)Birth Cohort 4*SC -0.002(0.016)Birth Cohort*OBC No No YesBirth Cohort Dummies No No YesReligion/Rural Dummies Yes Yes YesWoman’s Age Yes Yes YesState Fixed Effects No Yes YesObservations 107,373 107,373 107,373Mortality Neonatal Infant ChildScheduled Tribe 0.006** 0.009** 0.020***(0.003) (0.004) (0.005)Scheduled Caste 0.006*** 0.014*** 0.024***(0.002) (0.003) (0.003)OBC 0.006*** 0.008*** 0.011***(0.002) (0.002) (0.003)Child’s Year of Birth Fixed Effects Yes Yes YesMother’s Age at Birth Yes Yes YesFemale Child Dummy Yes Yes YesBirth Order Control Yes Yes YesIf Child was a Single Birth Yes Yes YesReligion/Rural Dummies Yes Yes YesState Fixed Effects Yes Yes YesObservations 184,087 184,087 184,087Note: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variabledescription is true and is 0 otherwise. Data source is DHS (2005-06) for the upper panel and DHS (1998-99)for the lower panel. Robust standard errors clustered at the state(district) level are in parentheses for theupper(lower) panel. Four (roughly) 10 year birth cohorts of the respondents have been computed for theupper panel; starting from year of birth 1948 upto 1991; later birth cohort numbers refer to younger women.Birth Cohort 1 is the omitted category.Data in the lower panel is at the respondent’s child level and restrictto all births at mother’s current place of residence. ***, ** and * indicate statistical significance at the 1%, 5% and 10% level of significance respectively. Regressions weighted by survey weight. Omitted categoryis non-SC/ST/OBC social group. Religion dummies included control for whether the respondent is Hindu,Muslim, Christian or Sikh; the omitted category being all other religions.903.9.TablesTable 3.5: Youngest Child’s Immunization RecordVariable Has Health-Card Any Vaccine BCG Measles All DPT All PolioST -0.196*** -0.123*** -0.156*** -0.153*** -0.172*** -0.029**(0.028) (0.027) (0.026) (0.022) (0.025) (0.012)SC -0.090*** -0.047** -0.084*** -0.083*** -0.087*** -0.016(0.020) (0.023) (0.027) (0.022) (0.022) (0.019)OBC -0.033** -0.002 -0.034 -0.034 -0.031 0.019(0.015) (0.026) (0.023) (0.020) (0.019) (0.037)Rural -0.138*** -0.072*** -0.115*** -0.105*** -0.124*** -0.137***(0.017) (0.019) (0.017) (0.013) (0.016) (0.012)Round 2005 0.041 0.158*** 0.068 0.110** 0.046 0.283***(0.060) (0.040) (0.051) (0.048) (0.054) (0.045)Round 2005*ST -0.040 0.041 0.010 -0.037 -0.038 -0.178***(0.061) (0.028) (0.040) (0.037) (0.042) (0.034)Round 2005*SC -0.024 0.003 -0.016 -0.032 -0.042 -0.094***(0.032) (0.025) (0.035) (0.035) (0.035) (0.029)Round 2005*OBC -0.049* -0.008 -0.024 -0.046 -0.056* -0.113**(0.026) (0.030) (0.028) (0.032) (0.032) (0.047)Religion Dummies Yes Yes Yes Yes Yes YesWoman’s Age Yes Yes Yes Yes Yes YesState Fixed Effects Yes Yes Yes Yes Yes YesObservations 61284 39516 61074 59878 60753 61194Note: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variable description is true and is 0 otherwise. Data source ispooled DHS (1998-99; 2005-06).Robust standard errors clustered at the state level are in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% level of significance respectively. Regressions weighted by survey weight. Omitted category is non-SC/ST/OBC social group. Religion dummiesincluded control for whether the respondent is Hindu, Muslim, Christian or Sikh; the omitted category being all other religions.913.9. TablesTable 3.6: Differences in Women’s EducationVariable Illiterate Years ofEducationST 0.347*** -2.947***(0.031) (0.241)SC 0.324*** -3.105***(0.023) (0.192)OBC 0.185*** -2.051***(0.028) (0.209)Birth Cohort 2*ST 0.034*** -0.467***(0.009) (0.108)Birth Cohort 3*ST 0.026** -0.781***(0.011) (0.135)Birth Cohort 4*ST -0.062* -0.498*(0.032) (0.262)Birth Cohort 2*SC 0.011 -0.358***(0.011) (0.116)Birth Cohort 3*SC -0.053*** -0.098(0.018) (0.163)Birth Cohort 4*SC -0.133*** 0.368(0.027) (0.277)Birth Cohort Dummies*OBC Yes YesBirth Cohort Dummies Yes YesReligion/Rural Dummies Yes YesWoman’s Age Yes YesState Fixed Effects Yes YesObservations 208,199 208,172Note: OLS estimations are reported. Outcome variable “illiterate” is a dummy variablethat assumes value 1 if the variable description is true and is 0 otherwise. Outcome vari-able “years of education” is a continuous variable. Data source is pooled DHS (1998-99;2005-06). Robust standard errors clustered at the state level in parentheses. ***, ** and *indicate statistical significance at the 1% , 5% and 10% level of significance respectively.Regressions weighted by survey weight. Omitted category is non-SC/ST/OBC socialgroup. Religion dummies included control for whether the respondent is Hindu, Muslim,Christian or Sikh; the omitted category being all other religions. Four (roughly) 10 yearbirth cohorts have been computed; starting from year of birth 1948 upto 1991; later birthcohort numbers refer to younger women. Birth Cohort 1 is the omitted category.923.9. TablesTable 3.7: Differences in Women’s Exposure to Amenities and MediaVariable: If Home Has Drinking Water Flush Toilet ElectricityST -0.304*** -0.159*** -0.228***(0.037) (0.022) (0.022)SC -0.215*** -0.136*** -0.147***(0.024) (0.026) (0.019)OBC -0.132*** -0.077*** -0.044(0.021) (0.026) (0.029)Round 2005 0.075 0.184*** 0.072(0.083) (0.056) (0.104)Round 2005*ST 0.277*** -0.098*** 0.027(0.055) (0.033) (0.044)Round 2005*SC 0.178*** -0.067* 0.008(0.030) (0.039) (0.035)Round 2005*OBC 0.111*** -0.062 -0.032(0.032) (0.038) (0.048)Observations 208215 208218 202900Variable If Watches TV If Listens to RadioST -0.260*** -0.150***(0.014) (0.014)SC -0.169*** -0.121***(0.010) (0.008)OBC -0.101*** -0.078***(0.010) (0.008)Observations 89577 89577Religion/Rural Dummies Yes Yes YesWoman’s Age Yes Yes YesState Fixed Effects Yes Yes YesNote: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variabledescription is true and is 0 otherwise. Data source is pooled DHS (1998-99; 2005-06) in top panel and DHS(1998-99) in bottom panel. Robust standard errors clustered at the state level in top panel and at district levelin bottom panel are in parentheses. ***, ** and * indicate statistical significance at the 1% , 5% and 10%level of significance respectively. Regressions weighted by survey weight. Omitted category is non-SC/ST/OBCsocial group. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian orSikh; the omitted category being all other religions.933.9.TablesTable 3.8: Differences in Women’s Status: Work and AutonomyVariable Currently Works Outside Spend Her What toWorking Home Wages CookScheduled Tribe 0.235*** 0.224*** 0.058*** 0.089***(0.026) (0.028) (0.008) (0.011)Scheduled Caste 0.169*** 0.176*** 0.058*** 0.051***(0.019) (0.019) (0.005) (0.007)OBC 0.099*** 0.073*** 0.018*** 0.015**(0.015) (0.016) (0.004) (0.006)Observations 89561 89577 89577 89577Variable Obtain Purchase Go to VisitHealthcare Jewellery Market Friends/RelativesScheduled Tribe 0.020** 0.021*** 0.046*** 0.036***(0.010) (0.006) (0.015) (0.014)Scheduled Caste 0.006 0.017*** 0.008 0.004(0.007) (0.004) (0.007) (0.006)OBC -0.001 0.005 -0.002 -0.004(0.006) (0.004) (0.007) (0.006)Observations 89577 89577 89577 89577Religion/Rural Dummies Yes Yes Yes YesWoman’s Age Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variable descriptionis true and is 0 otherwise. In particular, “Go to Market” and “Visit Friends/Relatives” assume value 1 if therespondent can do so without obtaining permission. All other outcome variables except those on work statusassume value 1 if the respondent can decide on them solely. Data source is DHS (1998-99). Robust standard errorsclustered at the district level in parentheses. ***, ** and * indicate statistical significance at the 1% , 5% and 10%level of significance respectively. Regressions weighted by survey weight. Omitted category is non-SC/ST/OBCsocial group. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh;the omitted category being all other religions. First two columns of top panel includes birth cohort and birth cohortinteracted with caste/tribe dummies. Seven (roughly) 5 year birth cohorts have been computed; starting from yearof birth 1948 upto 1985; later birth cohort numbers refer to younger women.943.9. TablesTable 3.9: Differences in Experiences at Health FacilityVariable Got Health Staff Spent Found FacilityService Sought Enough Time CleanScheduled Tribe -0.097*** -0.102*** -0.100***(0.015) (0.014) (0.014)Scheduled Caste -0.004 -0.013 -0.005(0.008) (0.008) (0.008)OBC -0.012 -0.015** -0.011(0.007) (0.007) (0.007)Observations 89577 89577 89577Religion/Rural Dummies Yes Yes YesWoman’s Age Yes Yes YesState Fixed Effects Yes Yes YesNote: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variabledescription is true and is 0 otherwise. Outcome variables refer to experiences of the respondents when shevisited a health facility/camp for herself or her children during the 12 months preceding the survey. Datasource is DHS (1998-99). Robust standard errors clustered at the district level in parentheses. ***, ** and* indicate statistical significance at the 1% , 5% and 10% level of significance respectively. Regressionsweighted by survey weight. Omitted category is non-SC/ST/OBC social group. Religion dummies includedcontrol for whether the respondent is Hindu, Muslim, Christian or Sikh; the omitted category being allother religions.953.9.TablesTable 3.10: Differences in Antenatal Care ReceivedVariable Doctor’s Professional Care Later No ANW Weight Blood Pressure Blood TestCare Care in Pregnancy Visit Measured Checked DoneST -0.204*** -0.160*** 0.590*** 0.165*** -0.122*** -0.169*** -0.136***(0.032) (0.028) (0.110) (0.027) (0.024) (0.031) (0.022)SC -0.130*** -0.094*** 0.535*** 0.092*** -0.055** -0.129*** -0.102***(0.017) (0.023) (0.053) (0.023) (0.025) (0.026) (0.023)OBC -0.076** -0.056** 0.264*** 0.055** -0.022 -0.043 -0.028(0.028) (0.024) (0.071) (0.024) (0.030) (0.038) (0.036)Round 2005 -0.029 0.028 -0.189* -0.046 0.078 -0.003 0.011(0.100) (0.092) (0.106) (0.093) (0.065) (0.072) (0.052)Round 2005*ST -0.094* -0.002 -0.051 -0.011 0.001 -0.051 -0.070*(0.052) (0.040) (0.142) (0.040) (0.042) (0.046) (0.039)Round 2005*SC -0.049 -0.006 -0.030 0.001 -0.064 -0.016 -0.044(0.033) (0.042) (0.065) (0.042) (0.039) (0.037) (0.030)Round 2005*OBC -0.040 -0.007 -0.036 0.007 -0.061 -0.046 -0.074(0.042) (0.049) (0.108) (0.050) (0.045) (0.048) (0.046)Religion/Rural Dummies Yes Yes Yes Yes Yes Yes YesWoman’s Age Yes Yes Yes Yes Yes Yes YesState Fixed Effects Yes Yes Yes Yes Yes Yes YesObservations 63779 63779 47821 63394 48095 48093 48083Note: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variable description is true and is 0 otherwise and pertain tolast pregnancy during three years preceding the survey. Data source is pooled DHS (1998-99; 2005-06). Robust standard errors clustered at the state level arein parentheses. ***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significance respectively. Regressions weighted by survey weight.Omitted category is non-SC/ST/OBC social group. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh; theomitted category being all other religions.963.9.TablesTable 3.11: Differences in Medical Care During DeliveryVariable: Assistance Doctor Any Traditional Birth Relative orFrom: Professional Attendant FriendST -0.191*** -0.232*** 0.087** 0.132***(0.026) (0.027) (0.037) (0.033)SC -0.139*** -0.131*** 0.053 0.070***(0.016) (0.018) (0.040) (0.020)OBC -0.073*** -0.062** -0.003 0.031(0.020) (0.028) (0.040) (0.037)Round 2005 0.067 0.057 0.088 -0.087(0.068) (0.070) (0.070) (0.057)Round 2005*ST -0.102** -0.088** 0.142** 0.047(0.042) (0.042) (0.064) (0.046)Round 2005*SC -0.035* -0.052* 0.053 0.053(0.020) (0.028) (0.050) (0.037)Round 2005*OBC -0.045 -0.052 0.100* 0.058(0.028) (0.040) (0.049) (0.050)Religion/Rural Dummies Yes Yes Yes YesWoman’s Age Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesObservations 63838 63838 63838 63838Note: OLS estimations are reported. Outcomes are dummy variables that assume value 1 if the variable description is trueand is 0 otherwise and pertain to last birth during three years preceding the survey. Data source is pooled DHS (1998-99;2005-06). Robust standard errors clustered at the state level are in parentheses. ***, ** and * indicate statistical significanceat the 1% , 5% and 10% level of significance respectively. Regressions weighted by survey weight. Omitted category is non-SC/ST/OBC social group. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian orSikh; the omitted category being all other religions.973.9.TablesTable 3.12: Knowledge of Modern Contraception: The Relative Contributions of Various FactorsSC STPredicted 0.98*** 0.93***(0.004) (0.019)Observations 18259 15041Explained By Attributes 0.05***(0.02)Unexplained By Attributes 0.005(0.007)Age in Years 0.0004* (0.0002)Rural Area 0.002 (0.001)Years of Education -0.002 (0.004)Contributions of Has TV 0.003* (0.001)Attributes Has Radio -0.00003 (0.0001)Drinking Water at Home 0.002* (0.001)Flush Toilet at Home 0.0001 (0.0002)Home Electrified 0.0003 (0.001)Family Planning Worker Visit 0.004** (0.002)Other Controls:Religion Dummy YesState FE YesNote: Data source is DHS (2005-06). Robust standard errors clustered at the state level are in parentheses. ***, ** and * indicatestatistical significance at the 1% , 5% and 10% level of significance respectively. Blinder-Oaxaca decomposition coefficients computedwith respect to a pooled model. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh;the omitted category being all other religions.983.9.TablesTable 3.13: Usage of Modern Contraception: The Relative Contributions of Various FactorsSC STPredicted 0.46*** 0.34***(0.02) (0.03)Observations 18259 15041Explained By Attributes 0.09***(0.03)Unexplained By Attributes 0.03***(0.01)Age in Years 0.004** (0.002)Rural Area 0.004* (0.002)Years of Education -0.0002 (0.0005)Contributions of Has TV 0.008*** (0.003)Attributes Has Radio 0.00002 (0.0001)Drinking Water at Home 0.002 (0.002)Flush Toilet at Home 0.001 (0.002)Home Electrified 0.002 (0.004)Family Planning Worker Visit 0.026** (0.011)Other Controls:Religion Dummy YesState FE YesNote: Data source is DHS (2005-06). Robust standard errors clustered at the state level are in parentheses. ***, ** and * indicatestatistical significance at the 1% , 5% and 10% level of significance respectively. Blinder-Oaxaca decomposition coefficients computedwith respect to a pooled model. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh;the omitted category being all other religions.993.9.TablesTable 3.14: Woman Got Tetanus Injection: The Relative Contributions of Various FactorsSC STPredicted 0.83*** 0.72***(0.04) (0.03)Observations 5861 5533Explained By Attributes 0.08**(0.04)Unexplained By Attributes 0.03***(0.01)Age in Years 0.003** (0.001)Rural Area -0.0001 (0.001)Years of Education -0.001 (0.004)Contributions of Has TV 0.002 (0.002)Attributes Has Radio -0.0001 (0.0002)Drinking Water at Home 0.004* (0.002)Flush Toilet at Home 0.0003 (0.001)Home Electrified 0.001 (0.003)Got Professional Prenatal Care 0.057** (0.025)Other Controls:Religion Dummy YesState FE YesNote: Data source is DHS (2005-06). Robust standard errors clustered at the state level are in parentheses. ***, ** and * indicatestatistical significance at the 1% , 5% and 10% level of significance respectively. Blinder-Oaxaca decomposition coefficients computedwith respect to a pooled model. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh;the omitted category being all other religions.1003.9.TablesTable 3.15: Woman Got Folic Tablets: The Relative Contributions of Various FactorsSC STPredicted 0.56*** 0.49***(0.02) (0.03)Observations 5263 4172Explained By Attributes 0.06*(0.03)Unexplained By Attributes 0.01(0.01)Age in Years 0.001* (0.0006)Rural Area 0.001 (0.001)Years of Education -0.004 (0.004)Contributions of Watches TV 0.002* (0.001)Attributes Listens to Radio -0.0003 (0.0004)Drinking Water at Home -0.0005 (0.002)No Toilet at Home -0.0005 (0.002)Home Electrified 0.001 (0.001)Got Professional Prenatal Care 0.021 (0.021)Other Controls:Religion Dummy YesState FE YesNote: Data source is DHS (1998-99). Robust standard errors clustered at the district level are in parentheses. ***, ** and * indicatestatistical significance at the 1% , 5% and 10% level of significance respectively. Blinder-Oaxaca decomposition coefficients computedwith respect to a pooled model. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh;the omitted category being all other religions.101Chapter 4The Legacy of FemaleLandlords in India4.1 IntroductionThere exists a large body of literature that analyses the role of historicalinstitutions, particularly property rights, on current outcomes (Acemogluet.al. (2001); Acemoglu and Johnson (2005); Banerjee and Iyer (2005)).However, few studies exist that explore the effect of historical female prop-erty rights on present development outcomes. This chapter, therefore, stud-ies the long term implications of women’s historical property rights. Inparticular, we study whether greater relative female landownership in colo-nial India influences outcomes in modern India. The current outcomes weconsider are literacy rate, infant mortality rate, indicators of healthcare forwomen, crimes against women, labour market participation as well as indi-cators of women’s autonomy.We consider female rent receivers as a proxy for female landownershipor property rights. The data on rent receivers are obtained from the Censusof India, 1921. The census of 1911 defines rent receivers as individualswho do not cultivate their land themselves, but sublet that land for rent.Thus, rent receivers would include all landowners who do not cultivate theirown land 21. The census of 1921 provides information on the number ofrent receivers in each district by gender. For the purpose of our analysis,we consider only those individuals for whom rent receiving was their mainor primary occupation and we exclude their dependants. In general, rent21Bengal Occupations, Census of India, 1911.1024.1. Introductionreceivers have been noted as not being workers and are, therefore, likelyto be owners of land 22. From that perspective, female rent receivers arelikely to represent women’s ownership of land and is, therefore, potentiallya measure of women’s property rights in the past.Our explanatory variable of interest is the ratio of female to male rentreceivers in a district in 1921. We match historic districts with their moderncounterparts using information from Banerjee and Iyer (2005) and officialdistrict websites that contain information on the district’s history. FollowingBanerjee and Iyer (2005) and Iyer (2010), we assign the same value of ourexplanatory variable of interest to modern districts if a historical districthas been eventually bifurcated to create these districts. The district-levelvariation in the proportion of female to male rent receivers appears to becorrelated with patterns of widowhood in the historic districts. Thus, his-toric patterns of male mortality and consequently widowhood for womenseem to be a plausible mechanism through which women became owners ofproperty. This also seems to be supported by historical accounts of femalelandlords who were predominantly widows.We find that greater fraction of female to male rent receivers in the pastis associated with greater literacy rate and lower infant mortality at present.Further, we find that higher relative historic female landownership is also as-sociated with greater labour force participation of women in particular, bet-ter provision of antenatal and postnatal healthcare, greater reporting of andarrests, chargesheeting for crimes against women as well as higher women’sautonomy in intra-household decision making. We then attempt to studythe possible mechanisms that could explain these findings. We consider po-litical and economic channels as likely mechanisms explaining our results.We find that greater relative female landownership in the past is associatedwith higher fraction of seats in state legislatures won by women. Further,greater investment in public goods such as in health, drinking water andpower are also found in districts with greater fraction of female to male rentreceivers. Also, districts with greater proportion of female to male rent re-22Change in the Work Force in British India, 1881-1921, Yoshifumi Usami (2006); SocialBackground of Indian Nationalism, A.R.Desai, 2005.1034.2. Dataceivers are found to have higher yields of rice and lower yields of wheat. Thisshows that districts with greater female landownership in the past are alsolikely to grow rice in which women have the potential to play an importantrole in cultivation; in contrast to wheat in which women’s role is limited.These findings suggest that greater relative property rights for women inthe past could influence current development outcomes that particularly af-fect women’s well-being plausibly through greater political representation ofwomen, investment in public goods as well as greater economic importanceof women in farming.This chapter proceeds as follows: section 4.2 describes the data and thesummary statistics; section 4.3 presents the empirical specifications; section4.4 presents the results; section 4.5 describes the plausible mechanisms thatcould explain the empirical findings and section 4.6 concludes.4.2 DataThe data used in this chapter come from a number of sources. For theanalysis in this chapter, we have used publicly available data from the Censusof India, 1921, 1981, 1991, 2011; Banerjee and Iyer (2005); Iyer et. al (2012);the Demographic and Health Surveys (DHS), 1998-99; and the NationalSample Survey (NSS), 2000. We describe the data sources as well as theexplanatory and outcome variables in the following sections.4.2.1 Data on Rent ReceiversThe explanatory variable of interest in this chapter is the proportion offemale to male rent receivers 23. The data on rent receivers have beenobtained from the Census of India, 1921. The imperial occupation tables inthe census provide the number of individuals who returned rent receivingas their primary occupation, their dependants as well as those for whom23We have also tried specifications where we have used the ratio of female to totalnumber of rent receivers in a district as our explanatory variable of interest. Our resultsremain unchanged even if we use this alternative measure of relative female landownershipas shown in Appendix C.1044.2. Datarent receiving is a subsidiary occupation. The number of rent receivers,categorized by gender, are available at the district level for 1921. For thepurpose of the present analysis, we consider only the number of individualsfor whom rent receiving was their primary occupation. From Table 4.1 wefind that the mean of the ratio of female to male rent receivers and femaleto total rent receivers at the district level is about 42 % (standard deviationof 59 %) and 26% (s.d. 12%) respectively. Figures 4.1 and 4.2 depict thedistrict-wise variation in our explanatory variables of interest. We find thatthere is considerable district-level variation in the fraction of female to maleand the fraction of female to total rent receivers in 1921 according to moderndistrict boundaries. We have also aggregated the number of female and malerent receivers and computed the ratio at the state level, when our outcomevariable is at the state level. Table 4.1 shows that the mean of female tomale and female to total rent receivers aggregated up to the state level isaround 38% (s.d. 12%) and 27% (s.d. 6%) respectively.4.2.2 Correlates of Historic Female LandownershipAn important question that might arise is what are associated with historicfemale landownership. Table 4.4 depicts the findings where we attempt tounderstand the factors that are correlated with relative female landowner-ship in the past. In Column (1) of Table 4.4 we regress the ratio of femaleto male rent receivers on relative female widowed population in 1921 alongwith geographic variables such as latitude, altitude, dummies for a district’ssoil type (such as indicators for black, alluvial and red soil), annual averagerainfall, indicator for whether the district is a coastal one as well as whetherthe district was under British rule. We find that higher proportion of femaleto male rent receivers is associated with higher fraction of female to malewidowed population historically from Column (1) 24. Column (2) includesan analogous specification as Column (1) and controls for the historic caste24We have also estimated a similar specification as Column (1) except that we have re-placed the relative female widowed population with the relative female married populationin 1921. We find that there does not appear to be any statistically significant correlationbetween the historic female landownership and relative female married population.1054.2. Datadistribution as well. In particular, even after controlling for the propor-tion of Brahmins in the Hindu population, historic female land ownershipappears to be positively correlated with historic relative female widowedpopulation (Column (2)). Column (3) also includes the historic proportionof district area under rice and wheat cultivation25. Our findings in Column(3) remain the same as in Columns (1) and (2). Column (4) includes theinteraction of historic relative female widowed population and proportionof Brahmins in the Hindu population. We find that historic relative femalelandownership appears to be higher in districts with greater relative femalewidowed population and greater fraction of Brahmins in the Hindu popu-lation. 26. Also, the historic area under rice or wheat does not appear tobe significantly associated with historic female landownership pattern as isseen in Table 4.4. Further, lower latitudes (that is, southern regions) appearto have a larger relative number of female to male rent receivers.Overall, widowhood or historic patterns of relative male mortality seemsto be a plausible mechanism through which women became owners of prop-erty. This is also supported by historical accounts of female landlords. RaniLakshmibai, Rani of Burdwan, Rani Bhabani were examples of female land-lords who were also widows with no or minor sons and managed their de-ceased husband’s estates. Although the British East India Company con-templated to disqualify female landlords from management and to bringtheir minor sons under company stewardship, such a policy was never imple-mented. During the colonial period, around one-third of men died withoutsurviving sons. A widow came to manage her husband’s property if she hadno male heirs or if she had minor sons/grandsons. Agarwal (2002) notesthat a widow could inherit her deceased husband’s estate in the absence ofmale heirs under both the Mitakshara and Dayabhaga schools of Hindu law.25The district’s area under rice/wheat at the time of independence is the earliest historicdata we have on area under different crops at the district level.26We have estimated a specification analogous to Column (4) but which also controlsfor historic proportion of Hindus and Muslims in the district population. The magnitudeand the sign on the interaction term between historic relative female widowed populationand proportion of Brahmins remains unchanged as in Column (4). We also do not findany association between female to male rent receivers and historic religious compositionof a district’s population.1064.2. DataHowever, the widow did not have the right to bequeath the land. On herdeath in the absence of her male heirs, the land would revert back to herhusband’s male relatives.4.2.3 Matching Historic Districts with Modern DistrictsThe analysis in this chapter focuses on the long term influence of relative fe-male landownership on current development outcomes. The ratio of femaleto male rent receivers that have been computed from the imperial tables ofthe census, 1921 are at the district level. Clearly, this variable correspondsto districts of British India in 1921. On the other hand, the current out-comes used in this chapter as outcome variables correspond to districts ofpost-independent India. Now, district boundaries have changed over time.However, comparing districts in 1921 with their counterparts in the post-independence period, we find that large changes in district boundaries haveoccurred mostly after 1947, that is, after independence 27. For the purposeof analysis in this chapter, we match current districts with their parent dis-tricts in 1921. We obtain information on the history of a district, namelythe history of its bifurcation, from the website maintained by the districtadministration. Further, we also studied the maps of the current districtsand the location of the historic districts, particularly for districts for whichthe district website could not provide adequate information that could ren-der matching possible. We also relied on Banerjee and Iyer (2005) for thepurpose of matching historic districts to their modern counterparts. Fol-lowing Banerjee and Iyer (2005) and Iyer (2010), we assign the same valueof relative female landownership to modern districts that were formed bybifurcating the same historic district.4.2.4 Data on Other ControlsA number of variables have been used as controls as they can potentiallyinfluence the outcome variables in this analysis. Our baseline empirical27See Mapping Indian Districts Across Census Years, 1971-2001, Kumar and So-manathan (2009) and Census of India 2011: Administrative Atlas of India.1074.2. Dataspecification includes geographic controls for latitude (in degrees), altitude(in metres), dummy variables for soil types (presence of black, alluvial andred soils), annual average rainfall (in millimetres), a dummy variable forwhether a district is a coastal one and whether the district was under theBritish rule. For empirical specifications where the outcome variables are atthe state level, the average of these variables across all districts in each ofthe states has been used in the analysis. The data on the geographic con-trols are obtained from Banerjee and Iyer (2005); district level reports of theCentral Ground Water Board (Ministry of Water Resource, Government ofIndia); district level reports of the Agriculture Contingency Plan (Depart-ment of Agriculture and Cooperation, Ministry of Agriculture, Governmentof India) and websites of the district administration for various districts.The summary statistics on control variables are provided in Table 4.1. Forindividual level outcome variables, we also include indicators for the indi-vidual’s caste/tribe, religion as well as whether she/he resides in rural areaand her/his age. We obtain this information from the NSS, 2000 and theDHS, 1998-99 that have been used for individual level outcome variables.In Appendix C, we present some results where we have included his-toric non-landlord proportion at the district level as a control. We obtainedthis measure from Banerjee and Iyer (2005). The objective is to analysewhether relative female landownership pattern continues to have any in-fluence on current outcomes even after controlling for colonial land tenuresystem as land tenure systems of British India have been documented toinfluence present outcomes (Banerjee and Iyer (2005)). Figure 4.3 plots thedifferent land tenure systems of British India by modern district boundaries.Landlord systems are found to be prevalent mostly in eastern India whilenon-landlord system are predominantly found in western and southern In-dia. Figure 4.4 plots the proportion non-landlord in British India by moderndistricts. Unlike Figure 4.3 that showed whether the district was under anon-landlord land tenure system or not, Figure 4.4 plots the proportion non-landlord prevalent in the district in the past. As before, districts in westernand southern India have a higher proportion that is under the non-landlord1084.2. Dataland tenure system than those in eastern India 28.4.2.5 Data on Outcome VariablesThe data on the outcome variables used in this chapter vary either at theindividual, district or state level and have been obtained from a number ofsources.Data on Literacy Rate: District and Individual LevelThe data on literacy rate at the district level have been obtained from Baner-jee and Iyer (2005) 29. The data on the literacy rate have been provided forthe years 1961, 1971, 1981 and 1991. The mean literacy rate in the sampleis about 30% (s.d. 12%) from Table 4.2.Further, we also include individual level educational attainment for womenand men from the National Sample Survey (NSS), 2000. From the NSSdataset, we construct a binary variable that takes the value 1 if an individ-ual is illiterate and zero otherwise. We also construct years of education foran individual. The NSS reports education codes that are converted to yearsof education following Hnatkovska et. al (2012) as follows: not literate=0years, literate but below primary= 2 years, primary=5 years, middle=8years, secondary and higher=10 years, graduate and above=15 years. Table4.3 reports the summary statistics on individual’s level of education. Thefraction of women who are illiterate is around 49% (s.d. 50 %) and theaverage number of years of education for women is around 3 years (s.d. 4years).28As the information on land tenure systems are not available for a number of districts,inclusion of proportion non-landlord reduces our sample size. Therefore, we present thoseresults in the appendix rather than in the main text. Also, as Banerjee and Iyer (2005)note that proportion non-landlord often does not vary within a state. So regressionsincluding proportion non-landlord do not include state fixed effects.29Appendix to Banerjee and Iyer (2005) describes that the literacy data have beenobtained from various censuses.1094.2. DataData on District Level Infant Mortality RateThe data on infant mortality rate at the district level are available fromBanerjee and Iyer (2005) 30 for the year 1991. Table 4.2 shows that theaverage infant mortality rate is around 83 per 1,000 births.Data on Health Outcomes: Individual LevelThe data on health outcomes at the individual level are for women only andhave been obtained from DHS (1998-99). We include variables that indicateprenatal and postnatal care obtained (such as obtaining prenatal care fromdoctor, receiving assistance from any professional medical personnel whilegiving birth, seeing doctor after giving birth, whether the child was weighedat birth), the quality of prenatal and postnatal care received (for instancegetting tetanus injection, folic tablets as well as getting abdomen checkedafter birth) and whether the respondent received advice from healthcareproviders regarding family planning, breastfeeding, baby care and whetherthe child was given Vitamin A tablets.Data on Crime: District and State LevelThe data on crime used in the chapter are both at the district and the statelevel. The district level data on crime are obtained from Banerjee and Iyer(2005). The authors have compiled data on crime rates at the district levelfor the years 1971, 1981 and 1991 from the district level reports on crimepublished by the National Crime Records Bureau, India. The district levelcrime variables that have been used in this analysis are the total numberof incidents of reported crimes, number of violent crimes 31, petty crimes32, number of incidents of stealing, number of events of dacoities and riots,number of attempted murders and number of incidents of women and menkidnapped.30See Appendix of Banerjee and Iyer (2005).31Violent crimes include murder, homicide, rape, dacoity, kidnapping,riots (Banerjeeand Iyer, 2005).32Petty crimes include cheating and counterfeiting (Banerjee and Iyer, 2005).1104.2. DataThe data on crime variables at the state level have been obtained fromIyer et. al (2012). The authors, in turn, have compiled the data from statelevel reports on crime incidence published by the National Crime RecordsBureau, India for the years 1985-2007. The crime variables have been cate-gorised according to whether they were specifically targeted against womenor not as well as variables on the arrests for crimes committed against menand women. Among the crimes specifically targeted against women, thischapter includes total number of reported incidents of crime against women,rapes, number of women kidnappings and murders. Among the crimes forwhich women are not the victims, we include total number of kidnappingsof men and boys, murders of men as well as property crimes and crimesagainst public order. This chapter also includes variables on the arrests forall crimes, rapes, crimes not committed against women, kidnappings of menand chargesheeting rates for all crimes as well as crimes against women.Table 4.2 presents the descriptive statistics on the district and state-levelcrime variables.Data on Labour Force Participation: District and IndividualLevelThe district level information on labour force participation have been ob-tained from the census, 2011. The variables included in this analysis arethe proportion of individuals in the rural sample who are main workers andmarginal workers by gender. The census of 2011 categorises workers intomain workers- those who have worked for at least 183 days in the year priorto the census and marginal workers- those who have not done so. The to-tal number of workers is the sum of main and marginal workers. Table4.2 reports that the fraction of women who are main workers and marginalworkers in the rural population is around 26% and 53% respectively.We use the individual level information on labour force participationfrom the Demographic and Health Surveys (DHS), 1998-99. The DHSdataset contains information at the individual level from both rural and ur-ban areas. We use the information contained in the DHS dataset on whether1114.3. Empirical Specificationthe respondent was currently working, works for someone else, earns cashfrom work and was employed during the last 12 months. This informationfrom the DHS is available for women between 15-49 years. From Table 4.3we find that, on an average, 37% women were currently working, 17% workfor someone else, 23% earn cash from work and 36% were employed duringthe last 12 months.Data on Women’s AutonomyVariables on women’s autonomy are obtained from the DHS, 1998-99. Inparticular, these variables attempt to capture women’s decision-making abil-ity in the household and mobility. The autonomy variables are binary vari-ables that assume the value of 1 if the respondent can solely decide whatto cook, whether to seek healthcare for herself, to purchase jewellery, tostay with family and how to spend the wages earned from work. We alsoinclude other binary variables that take the value of 1 if the woman does notrequire permission to go to the market, visit friends/relatives and whethershe was not beaten since the age of 15. We also construct autonomy in-dices comprising of the aforementioned variables that indicate autonomyand mobility. The higher the values of these indices, the higher is the levelof woman’s autonomy. Table 4.3 provides the descriptive statistics on theseautonomy variables. On an average, 70% women can solely decide what tocook, 28% can decide to obtain healthcare, 10% can purchase jewellery, 13%can decide whether to stay with family, 10% can decide how to spend thewages earned from work, 32% do not require permission to go to the market,25% do not require permission to visit their friends and relatives and 80%report not been beaten since the age of 15.4.3 Empirical SpecificationIn this section, we outline the empirical specifications used in this chapter.As our outcome variables vary at the district, state and individual levels, wepresent different empirical specifications depending on the level of variation1124.3. Empirical Specificationof the outcome variable included in that specification.4.3.1 Empirical Specification: District Level OutcomeVariablesThe empirical specification for district level outcome variables is as follows:ydst = α+ βfemaletomaleRRd1921 + γXds + φs + ψt + εdst (4.1)Here, ydst refers to the outcome variable of interest in district d of state sin the year t. In our analysis, the outcome variables that vary at the districtlevel are some crime variables, literacy rates, infant mortality rate and labourforce participation rates. Also, here femaletomaleRRd1921 refers to the ratioof female to male rent receivers in district d (of state s) in 1921, where thedistrict d has been matched to its historic counterpart. Xds refers to timeinvariant geographic and historical variables that are potentially exogenousand can likely influence the outcome variable of interest. In our analysis,we include altitude, latitude, soil characteristics (whether the district hasred, black or alluvial soil)and annual average rainfall of the district as wellas whether the district is a coastal district as the exogenous geographiccontrols. We also include a dummy for whether the district was under theBritish. φs and ψt are the state and year fixed effects respectively; the latterincluded when data are available on the outcome variable for multiple years.εdst is the regression disturbance term that is clustered at the district level.4.3.2 Empirical Specification: State Level OutcomeVariablesThe empirical specification that we use for state level outcome variables isas follows:yst = α+ βfemaletomaleRRs1921 + γXs + φR + ψt + εst (4.2)1134.3. Empirical SpecificationHere, yst refers to the outcome variable of interest of state s in the yeart. In our analysis, the outcome variables that vary at the state level aresome crime rates.femaletomaleRRs1921 refers to the ratio of female to malerent receivers in state s in 1921. The data on rent receivers available at thedistrict level have been aggregated up to the state level in this case. Xsrefers to time invariant geographic and historical variables for state s thatare potentially exogenous and can likely influence the outcome variable ofinterest. We include the altitude, latitude, annual average rainfall of a stateas some of the geographic controls in Xs and these variables are obtainedby computing the average of these variables across all districts in the state.Other geographic controls included in Xs are the proportion of districts in astate that have black, alluvial and red soils as well as the fraction of districtsin a state that are coastal. We compute the fraction of districts in a statethat were under the British and include it as the historical control variablein Xs. φR is the region fixed effects and ψt is the year fixed effect. εst is theregression disturbance term that is clustered at the state level.4.3.3 Empirical Specification: Individual Level OutcomeVariablesThe empirical specification that we use for individual level outcome variablesis as follows:yids = α+ βfemaletomaleRRd1921 + γXds + δZids + φs + εids (4.3)Here, yids refers to the outcome variable of interest for individual i indistrict d of state s. The outcome variables that vary at the individual levelin this chapter are education, labour market participation, health outcomesand autonomy of women. The individual level outcome variables are avail-able only at a given point in time. As before, femaletomaleRRd1921 refersto the ratio of female to male rent receivers in district d in 1921. Time-invariant geographic and historic variables corresponding to district d instate s are altitude, latitude, soil characteristics (whether the district has1144.4. Resultsred, black or alluvial soil)and annual average rainfall of the district as wellas whether the district is a coastal one and dummy for whether the districtwas under the British. Zids is the vector of individual level controls andincludes the age, dummies for caste/tribe and religion of the individual ias well as dummy variable for whether the individual i resides in the ruralarea. φs refers to state fixed effects and εids is the regression disturbanceterm that is clustered at the district level.4.4 ResultsWe present our empirical results in this section. We categorise our resultsaccording to the outcome variables. At first we present the results on theliteracy outcome variables, followed by those on infant mortality rate, healthoutcomes, crime rates, labour force participation and lastly on women’sautonomy.4.4.1 Literacy VariablesTable 4.5 presents the results on the association between literacy rate atthe district level and the ratio of female to male rent receivers in the up-permost panel. We focus on Column (3) for the purpose of interpretationof the results as it includes all possible controls as well as year and statefixed effects. One standard deviation increase in the relative fraction of fe-male to male rent receivers is associated with 1.5 percentage points highercurrent district level literacy rate. Appendix Table C.7 runs an analogousspecification as Column (3) of Table 4.5, but includes the ratio of femaleto total rent receivers as the explanatory variable of interest instead. Ourresults remain unchanged even if we include this alternative measure of rel-ative female landownership in the past. In Column (4) of Table 4.5, we usethe same regression specification as Banerjee and Iyer (2005) and includedistrict level proportion non-landlord as an additional explanatory variable.We find that despite including the historic non-landlord proportion, higherproportion of female to male rent receivers is associated with higher current1154.4. Resultsliteracy rate.The middle and lower panels of Table 4.5 report results on the probabilityof an individual being illiterate from the NSS, 2000 for women and menrespectively. We control for the respondent’s caste/tribe, religion, age aswell as whether she resides in rural area in our regression specification forthese individual level outcome variables. Column (3) of the middle and lowerpanels of Table 4.5 show that one standard deviation increase in the relativefemale to male rent receivers is associated with around 1.5 percentage lowerlikelihood of being illiterate for both women and men. We obtain similarfindings when we use the proportion of female to total rent receivers as ourexplanatory variable of interest as shown in Appendix Table C.7. Our resultsremain robust to controlling for the measure of proportion non-landlord inthe district following Banerjee and Iyer (2005) (Column (4) in the middleand lower panels of Table 4.5) 33.4.4.2 Infant Mortality and Health OutcomesTable 4.6 reports the results where different outcome variables that are indi-cators of health are presented. We use the regression specification of Column(3) of Table 4.5 for our analysis. In particular, for district level outcome vari-able, the regression specification of Column (3) in the uppermost panel ofTable 4.5 has been used. For individual level outcome variables, our regres-sion specification is analogous to Column (3) of the middle and lower panelsof Table 4.5 that also include demographic controls. Table 4.6 shows thatone standard deviation increase in the fraction of female to male rent re-ceivers in the past is associated with a decrease in infant mortality by around3 deaths. This is also robust to including Banerjee-Iyer non-landlord pro-portion as a control as shown in Appendix Table C.2. Appendix Table C.7shows that when we use the ratio of female to total rent receivers as our ex-planatory variable of interest, the coefficient on infant mortality is negative,but not statistically significant.33Although not presented here, we do not find any significant association between his-toric relative female landownership and years of education for women.1164.4. ResultsWe consider different aspects of health for women as outcome variablesusing the DHS, 1998-99 data on ever-married women aged 15-49 years. Weconsider the quality of pre and postnatal care and health related advicereceived by women from healthcare providers after giving birth as our out-comes of interest. It is to be noted that we control for the respondent’scaste/tribe, religion, age as well as whether she resides in rural area in ourregression specification for these individual level outcome variables.Table 4.6 deals with outcomes such as whether the woman respondentreceived prenatal care from a doctor, assistance from a professional health-care provider (doctor, trained nurse) at delivery, the child was weighed atbirth, got tetanus and folic tablets during her pregnancy, she saw a doctorafter giving birth, whether her abdomen was examined as part of postnatalcheck-up and if her last-born child was given Vitamin A tablets to protecther/him from night-blindness. The information on prenatal, postnatal andcare at delivery are available for the last child born to the respondent withinthe three years preceding the survey. We find that one standard deviationincrease in the fraction of female to male rent receivers in the past is asso-ciated with increased likelihood that the respondent got prenatal care froma doctor by 2.2 percentage points, saw a doctor within 2-3 months of givingbirth by 2.7 percentage points and got her abdomen examined (postnatal)by 2.4 percentage points. Further, the likelihood of not receiving any assis-tance from a professional healthcare provider at the time of delivery (suchas doctor, trained nurses, community health worker) is found to be lower by4 percentage points in districts that had a higher relative ratio of female tomale rent receivers historically. Also, higher relative female landownershipis associated with lower likelihood of the child not being weighed at birth by4.2 percentage points. We do not however find any significant associationbetween historic female landownership patterns and getting tetanus injec-tion or folic tablets for consumption as part of antenatal care. Also, we findthat a one standard deviation higher fraction of female to male rent receiversin 1921 is associated with an increased probability that the respondent af-ter giving birth received advice about family planning by 2.3 percentagepoints, breastfeeding by 3.5 percentage points, how to care for the baby by1174.4. Results2.7 percentage points. Further, the last born child is more likely to receiveVitamin A tablets from healthcare providers by 2.3 percentage points. Weobtain similar findings when use the ratio of female to total rent receiversas the explanatory variable of interest as is shown in Appendix Table C.7.4.4.3 CrimeTable 4.7 includes outcome variables on reported crimes against women,those not specifically targeted against women and indicators of law enforce-ment like chargesheeting and arrests at the state level. We find that totalreported crimes against women, reported incidents of rape and murders ofwomen appear to be higher in states that had greater proportion of femaleto male rent receivers. We also find that reported incidents of kidnappingsof men and boys are negatively and reported incidents of property crimes arepositively associated with historical state level ratio of female to male rentreceivers. We study whether higher reported incidents of crimes, especiallyagainst women, is associated with better law enforcement in states whichhad higher relative female landownership historically. From Table 4.7 wefind that in states with higher ratio of female to male rent receivers, arrestsfor crimes against women and rapes appear to be higher. Also, arrests forkidnappings against men appear to be lower in states with higher relativefemale landownership in the past. Further, Table 4.7 shows that chargesheet-ing rate for overall crimes do not appear to be associated with female rentreceivers. However, chargesheeting rate for crimes against women are foundto be higher in states with greater fraction of female to male rent receiversin the past. Further, our results are robust to including the historical non-landlord proportion from Banerjee and Iyer (2005) as shown in AppendixTable C.3. We find similar results if we use the ratio of female to totalrent receivers at the state level instead as our main explanatory variable asAppendix Table C.9 depicts.Appendix Table C.1 reports district level crime variables obtained fromBanerjee and Iyer (2005). Reporting of total crimes as well as reported inci-dents of petty and stealing crimes appear to be larger in districts with higher1184.4. Resultsrelative ratio of female to male rent receivers in the past. These findingsare also robust to including the proportion non-landlord as an additionalcontrol as shown in Appendix Table C.2 34.4.4.4 Women’s Labour Market ParticipationWe consider the association between women’s labour market participation atboth the district and individual levels and historical relative female landown-ership. We also consider whether there has been any influence on labourmarket participation of men.Table 4.8 shows that one standard deviation increase in female to malerent receivers is associated with an increase in the fraction of women mainworkers in the rural population by 1 percentage point (main workers arethose who have worked for more than 183 days in the year preceding thecensus year, 2011). We find no significant effect on the proportion of femalemarginal workers in the rural population (marginal workers work for lessthan 183 days in the year before the survey) 35. Also, we do not find anysignificant influence on men’s labour market participation at the districtlevel. Our results remain unaffected even after we control for proportionnon-landlord in the district in the past as shown in Appendix Table C.2.We find almost analogous results when we use the proportion of female tototal rent receivers at the district level as Appendix Table C.8 depicts.Table 4.8 presents some additional results on women’s employment usingthe DHS, 1998-99 data. We find from Table 4.8 that higher relative femaleto male rent receivers in the past is associated with greater likelihood thatwomen in those districts were currently working, working for someone elsewho is not a family member, earning cash from work and are also more likely34State level crime variables are more detailed in terms of specifying in detail whetherthey are specifically against women or not as well as law enforcement measures. Hencewe include them in the main text and present the district level outcomes that are not asdetailed in the Appendix.35Although not presented in Table 4.8, we find that higher female to male rent receiversin the past is associated with an overall increase in the fraction of women workers in therural population. It is, therefore, likely that the overall higher incidence of female workersin the rural population is on account of higher fraction of women main workers in therural population in districts that had a higher relative female landownership in the past.1194.4. Resultsto be employed during the last 12 months. Specifically from Table 4.8 wefind that one standard deviation higher proportion of female to male rentreceivers at the district level is associated with an increase in the probabil-ity that a woman living in the district would be currently working by 3.2percentage points, working for someone who is not a family member by 2.3percentage points, earning cash by 3 percentage points and was employedduring the last 12 months by 3.2 percentage points. Appendix Table C.4finds that these results are likely to hold for women residing in rural areas.Our results are unaffected if we include the proportion of female to total rentreceivers instead as our outcome variable of interest as is shown in AppendixTable C.8.Therefore, we find that women living in districts which had higher frac-tion of female to male or female to total rent receivers in the past are morelikely to participate in the labour market. Further, women are also morelikely to be working for someone else and earning cash from work. Thisfinding is particularly important because it can play a major role in enhanc-ing women’s autonomy as demonstrated by Anderson and Eswaran (2009),who show that rather than being employed, it is working outside one’s hus-band’s farm and therefore earning income that can contribute to women’sempowerment.4.4.5 Women’s AutonomyWe study whether historical female property rights can influence women’sautonomy currently using individual level data from the DHS, 1998-99 col-lected from ever-married women between the ages of 15 and 49 years. Inparticular, we study outcomes that are related to women’s participationin intra-household decision making, freedom of mobility and exposure tophysical abuse. We consider whether the respondent can solely decide whatto cook, obtain healthcare for herself, purchase jewellery, stay with familyand how to spend wages she earned from work as outcomes that pertainto women’s ability in making decisions in the household. We also considerwhether the respondent can go to the market and visit her friends or relatives1204.4. Resultswithout seeking permission as indicators of women’s freedom of mobility. Weinclude a dummy variable that assumes the value 1 if the respondent wasbeaten since the age of 15 and is 0 otherwise as an indicator for women’slikelihood of facing physical violence.Table 4.9 shows that one standard deviation higher fraction of femaleto male rent receivers at the district level in the past is found to increasethe likelihood that the respondent can solely decide to obtain healthcare forherself by 3.2 percentage points, purchase jewellery by 1.4 percentage points,stay with family and how to spend her wages by 1.6 percentage points as wellas visit her friends/relatives without seeking permission by 2.5 percentagepoints. Although the coefficient on the ratio of female to male rent receiversfor the outcome whether a respondent was beaten since she was 15 years oldis negative, it is not statistically significant. We obtain analogous findingswhen we consider the ratio of female to total rent receivers as our explanatoryvariable of interest as can be seen from Appendix Table C.8.In Appendix Table C.5 we construct four “indices of autonomy” as ouroutcome variables. These autonomy indices combine the different outcomescontained in Table 4.9 to measure overall autonomy. Specifically, Auton-omy Index 1 has five components- whether the respondent can solely decidewhat to cook, obtain healthcare, purchase jewellery, stay with family andhow to spend her wages. The value of the index increases by 1 when therespondent is solely able to decide for herself on each of these situations.Thus, the maximum value of Autonomy Index 1 is 5 which indicates thatthe respondent can solely decide on each of these situations and assumesthe minimum value 0 when she is unable to decide on any of these situa-tions on her own. Therefore higher the value of Autonomy Index 1, higheris the respondent’s autonomy. We construct the other autonomy indicesanalogously. Autonomy Index 2 includes all components of Autonomy In-dex 1 but excludes whether the respondent can decide what to cook andpurchase jewellery on her own. This is because women are less likely to faceimpediments from household members in deciding what to cook and buyjewellery by themselves. Autonomy Index 2 assumes the maximum value 3and the minimum value 0. Autonomy Index 3 includes the components of1214.5. Possible MechanismsAutonomy Index 1 as well as whether she can go to the market, visit herfriends/relatives without obtaining permission and if she is allowed to havemoney set aside for her use. Its maximum value is 8 and minimum valueis 0. Autonomy Index 4 consists of the components of Autonomy Index2 along with whether she can go to the market, visit her friends/relativeswithout obtaining permission and if she is allowed to have money set aside.The highest value of this index is 6 and the lowest value is 0.Appendix Table C.5 shows that one standard deviation increase in theratio of female to male rent receivers in the past is associated with an increasein Autonomy Index 2 by 4.8 percentage points, Autonomy Index 3 by 9.6percentage points and Autonomy Index 4 by 9 percentage points.4.5 Possible MechanismsHere, we study the potential channels through which historic female propertyrights can have long term influence on development outcomes. We considerthe influence of women politicians, public good provision and women’s con-tribution in agriculture as plausible channels through which historic femalelandownership patterns can have long term effects.4.5.1 Women Politicians and Public GoodsWe consider whether women politicians is a plausible mechanism throughwhich historic female property rights influences long-term outcomes. Thisis because a large body of literature exists which document that individualsborn in districts with women politicians are more educated (Clots-Figueras(2012)) and such districts also appear to have lower neonatal and infantmortality on account of greater investment in public health infrastructure bywomen politicians (Bhalotra and Clots-Figueras (2014)). Further, Beamanet. al. (2009) show that prior exposure to women politicians changes voterattitudes and likely weakens gender stereotypes. Beaman et. al. (2012)also show that women politicians act as role models and help in reducinggender gap in educational attainment. Therefore, we are motivated to anal-1224.5. Possible Mechanismsyse whether districts with greater fraction of female to male rent receiversin the past are also more likely to have greater proportion of seats in statelegislatures won by women politicians.The first column in Panel A of Table 4.10 shows that higher proportionof female to male rent receivers in the past is associated with higher fractionof seats in the state legislatures being won by women politicians.Since we found that greater historic female property rights is associ-ated with greater fraction of seats won by women politicians, we investigatewhether provision of public goods is also higher currently. Using VillageAmenities Data from Census of India, 1991 (obtained from Banerjee andIyer (2005)), we investigate the effect of female to male rent receivers on theproportion of villages in a district having different public goods and healthseeking behaviour at the individual level in Table 4.10.From Table 4.10 we find that although districts with greater proportionof female to male rent receivers in the past is unlikely to have larger frac-tion of villages having educational institutions (such as primary and highschools), but they appear to be associated with greater proportion of villagesthat have health facilities (such as dispensary and health centre), drinkingwater facility such as taps and power for domestic uses. Further, we also findthat women in districts that had a greater relative female landownership inthe past are less likely to give birth at home without any medical assistance.Therefore, greater relative property rights for women historically couldplausibly influence current outcomes through better provision of public goodsin those districts at present.4.5.2 Women’s Contribution in AgricultureWe consider the role of women’s contribution in agriculture as a plausiblemechanism through which female rent receivers may influence long termdevelopment outcomes. This is because the literature has documented thatincrease in the relative importance of women in agriculture can improve sur-vival rate and educational attainment of girls (Qian, 2008) and use of certainagricultural implements that require significant upper body strength like the1234.5. Possible Mechanismsplough, raises the relative importance of men in agriculture; thereby result-ing in unequal gender norms (Alesina et. al, 2013). In the current context,we consider whether districts that had a higher proportion of female to malerent receivers in the past are more likely to grow rice or wheat as women aremore likely to have a greater contribution in the cultivation of rice relativeto that of wheat. Using data from Banerjee and Iyer (2005) on log of agri-cultural yields at the district level for the period 1956-1987, we find fromPanel A of Table 4.11 that districts which had higher relative number of fe-male to male rent receivers in the past not only report higher yields of majorcrops; but are found to have higher rice yields and lower wheat yields as well.In Panel B of Table 4.11, we also control for investments in farming thatcould account for crop yields such as proportion of gross cropped area thatis irrigated, fertilizer use, proportion of crop area under HYV (high-yieldingvariety) variety of that crop in addition to historic female landownershippattern. We find that our results remain unaffected even after we controlfor agricultural investments that could affect agricultural yields. As a ro-bustness check, we run the same regression specifications as in Table 4.11,but by also including Banerjee and Iyer (2005) variable on proportion non-landlord and report these findings in Appendix Table C.6. This is to seethat whether our explanatory variable of interest on female landownershipcontinues to have any influence on agricultural yields even after we controlfor historical land tenure system in the district. We find that districts thathad a higher ratio of female to male rent receivers in the past are likely tohave lower wheat yields and this continues to hold even after we control foragricultural investments as in Table 4.11. Therefore we find that women’scontribution in agriculture is higher in districts that had greater propertyrights for women historically, in the form of higher yield of rice and loweryield of wheat. Thus, historic female property rights could influence longterm development outcomes, particularly pertaining to women, plausiblythrough greater role of women in agriculture.1244.6. Conclusion4.6 ConclusionIn this chapter we study the long term implications of historical femaleproperty rights on current development outcomes. We consider the ratio offemale to male rent receivers in Indian districts in 1921 as a potential indi-cator of historic relative female property rights and study its implicationson current outcomes. These include literacy rate for women and men, infantmortality, quality of pre and postnatal healthcare received by women, advicereceived by women from healthcare providers, reported crimes and arrestsfor crimes against women, women’s and men’s labour market participationand women’s autonomy. We find that historic patterns of widowhood forwomen and therefore of male mortality is a plausible mechanism throughwhich women became landowners. However, historic crop choices do notseem to influence relative female landownership pattern in the past. We findthat greater proportion of female to male rent receivers in the past is asso-ciated with better literacy outcomes for both men and women, lower infantmortality, better healthcare received by women, greater reported crimes andarrests for crimes committed against women, higher labour market partici-pation only for women and better women’s autonomy. Further, we obtainsimilar findings if we use the ratio of female to total rent receivers as analternative explanatory variable of interest. We investigate the plausiblemechanisms through which historical property rights for women can influ-ence current development outcomes. We observe that districts that had ahigher relative ratio of female to male rent receivers are also more likely tohave women politicians in state legislatures as well as higher provision ofsome public goods. Also, women’s contribution in agriculture is found tobe higher as rice yields are found to be higher relative to wheat yields indistricts with greater female relative landownership in the past. Therefore,women politicians, investment in public goods and greater role of women inagriculture appear to be potential channels through which property rightsfor women in the past continue to have long term effects.1254.7. Figures4.7 Figures¯Female to Male Rent Receivers, 1921 by Modern BoundariesFemale to Male Rent Receivers, 1921Data Unavailable0.00 - 0.200.21 - 0.400.41 - 0.600.61 - 0.800.81 - 1.001.01 - 1.50>1.50Figure 4.1: Female to Male Rent Receivers, 19211264.7. Figures¯Female to Total Rent Receivers, 1921 by Modern BoundariesFemale to Total Rent Receivers, 1921Data Unavailable0.00 - 0.100.11 - 0.200.21 - 0.300.31 - 0.400.41 - 0.600.61 - 0.800.81 - 1.00Figure 4.2: Female to Total Rent Receivers, 19211274.7. FiguresHistoric Land Revenue SystemData UnavailableOnly Landlord Non-LandlordLand Revenue Systems of British India by Modern Boundaries¯Figure 4.3: Land Revenue Systems of British India, (Banerjee & Iyer, 2005)1284.7. FiguresHistoric Proportion Non-LandlordData Unavailable0.000.01 - 0.200.21 - 0.400.41 - 0.600.61 - 0.800.81 - 0.991.00District-wise Historic Proportion Non-Landlord  by Modern Boundaries¯Figure 4.4: Proportion Non-Landlord in British India, (Banerjee & Iyer,2005)1294.8. Tables4.8 TablesTable 4.1: Descriptive statistics of Explanatory VariablesVariable Mean Std. Dev. ObservationsPanel A:District Level:Total Rent Receivers 12605.63 24866.94 457Male Rent Receivers 8749.86 15483.62 459Female Rent Receivers 3817.55 9843.58 457Female RR/Male RR 0.42 0.59 457Female RR/Total RR 0.26 0.12 457Panel B:State Level:Total Rent Receivers 11734.13 15551.35 18Male Rent Receivers 8124.14 9807.65 18Female Rent Receivers 3601.32 5810.04 18Female RR/Male RR 0.38 0.12 18Female RR/Total RR 0.27 0.06 18Panel C:District Level:Altitude 434.67 663.81 451Latitude 22.80 5.96 451Black Soil 0.24 0.42 451Alluvial Soil 0.41 0.49 451Red Soil 0.23 0.42 451Annual Average Rainfall 1214.04 605.57 451Coastal District 0.13 0.33 452Year of British Land Revenue Control 1802.84 30.26 277British Ruled District 0.64 0.48 460Note: Data source is the imperial table, Census of India, 1921 for Panels A and B and Banerjeeand Iyer (2005) for Panel C. For the Panel B, the mean and the standard deviation for each ofthe variables have been computed by aggregating the district level variables up to the state level.For Panel C, all variables except “Annual average rainfall” and “Year of British Land RevenueControl” are dummy variables that assume the value 1 if the description in the variable name istrue and is 0 otherwise.1304.8. TablesTable 4.2: Descriptive statistics of Outcome Variables IVariable Mean Std. Dev. ObservationsDistrict Level:Literacy Rate 0.30 0.12 991Infant Mortality 83.11 29.33 270Female Workers-Proportion 0.34 0.10 387Female Main Workers-Proportion 0.26 0.11 991Female Marginal Workers-Proportion 0.53 0.12 387Total Crime/1000 pop 2.02 1.28 640Violent Crime/1000 pop 0.26 0.16 641Petty Crime/1000 pop 0.06 0.05 641Stealing Crime/1000 pop 0.82 0.69 640Dacoity/ 1000 pop 0.16 0.12 637Attempted Murder/1000 pop 67.63 84.52 265Women Kidnapped/1000 pop 30.13 26.80 265Others Kidnapped/1000 pop 12.35 27.05 265State Level:Crimes against Women/1000 women 0.17 0.14 391Rapes/1000 women 0.03 0.02 391Kidnap women-girls/1000 women 0.04 0.03 340Murders of Women 452.44 349.02 136Kidnap Men-Boys/1000 men 0.02 0.01 340Murders of Men 1556.68 1442.54 136Property Crimes/1000 pop 0.45 0.19 391Public Order Crimes/1000 pop 0.11 0.09 391Arrests Woman Crimes/ 1000 pop 0.31 0.23 391Arrests Rape/1000 pop 0.02 0.01 391Arrests Non-Woman Crimes/1000 pop 0.01 0.005 391Arrests Man Kidnaps/1000 pop 0.01 0.01 391Chargesheet Rate-All Crimes 76.66 10.91 289Chargesheeting Rate-Woman Crimes 89.60 7.75 289Note: Data Source is Banerjee and Iyer (2005) for literacy, infant mortality and crime variables at thedistrict level; Census of India, 2011 for labour market participation variables; Iyer et. al (2012) for statelevel crime variables . Mean and standard deviations have been reported in this table.1314.8. TablesTable 4.3: Descriptive statistics of Outcome Variables IIVariable Mean Std. Dev. ObservationsPanel A:Years of Education 3.15 4.12 253211If Illiterate 0.49 0.50 256314Panel B:If No Work Currently 0.63 0.48 68657If Works For Someone Else 0.17 0.38 68657If Earns Cash 0.23 0.42 68657If Worked Last Year 0.37 0.48 68657Can Herself Decide How to Spend Wages 0.10 0.30 68657Can Herself Decide What to Cook 0.70 0.46 68650Can Herself Decide to Obtain Healthcare 0.28 0.45 68648Can Herself Decide to Buy Jewellery 0.10 0.30 68646Can Herself Decide to Stay With Family 0.13 0.33 68627No Permission Needed to Go to Market 0.32 0.47 68640No Permission Needed to Visit Relatives 0.25 0.43 68641Not Beaten Since Age 15 0.80 0.40 68652Note: Data source is the Employment-Unemployment Survey (Schedule 10) of the National Sample SurveyOrganization (NSSO), 2000 for Panel A and Demographic and Health Surveys (DHS), 1998-99 (Women’sQuestionnaire) for Panel B. Mean and standard deviations have been reported in this table. Years ofeducation is a continuous variable. All other variables are binary variables. All variables are reported forwomen in the sample.1324.8.TablesTable 4.4: Correlates of Historic Female Landownership(1) (2) (3) (4)Female RR/Male RRFemale/Male Widowed 0.041** 0.048** 0.039* -0.007(0.019) (0.021) (0.022) (0.020)Proportion Brahmin 0.163 0.242 -1.052(0.157) (0.570) (0.603)Female/Male Widowed*Prop Brahmin 0.511**(0.226)Proportion Rice Area -0.086 -0.028(0.126) (0.130)Proportion Wheat Area -0.249 -0.392(0.221) (0.319)Latitude -0.030* -0.032** -0.028* -0.030**(0.015) (0.015) (0.015) (0.014)British Ruled 0.135 0.127 0.048 0.152(0.079) (0.081) (0.077) (0.088)R-squared 0.207 0.209 0.223 0.229Observations 416 407 388 388Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Data source is Census of India, 1921; Indian Agricultural Statistics 1947-1949 and Banerjee andIyer (2005). Observations are at the district level. Robust standard errors clustered at the modern statelevel are in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% level ofsignificance respectively. “Proportion Rice/Wheat Area” refers to the fraction of a district’s area underrice/wheat at the time of independence. “Geo Controls” include altitude, latitude, whether the districthas black, red or alluvial soil, if the district is coastal and annual average rainfall. Altitude refers to theaverage altitude of the district in metres, Latitude refers to latitude of the centre of the district/districtheadquarters in degrees. Black soil, alluvial soil and red soil indicate dummy variables that assume thevalue 1 if the district has black soil, alluvial soil and red soil respectively; assuming the value 0 otherwise.Coastal district is a dummy variable that assumes the value 1 if the district is a coastal one and is 0otherwise.1334.8. TablesTable 4.5: Literacy RateDistrict Level: (1) (2) (3) (4)District Level:Female RR/Male RR 0.022*** 0.019*** 0.026*** 0.020***(0.004) (0.003) (0.003) (0.003)British Ruled 0.038*** 0.039***(0.011) (0.012)Proportion Non-Landlord 0.057***(0.019)Year Fixed Effects Yes Yes Yes YesR-squared 0.535 0.555 0.714 0.572Observations 972 972 972 583If Illiterate: Females (1) (2) (3) (4)Female RR/Male RR -0.060** -0.063** -0.025** -0.039**(0.025) (0.026) (0.012) (0.016)British Ruled 0.014 -0.042***(0.014) (0.013)Proportion Non-Landlord -0.041**(0.020)Demographic Controls Yes Yes Yes YesR-squared 0.175 0.175 0.197 0.170Observations 244,696 244,696 244,696 158,920If Illiterate: Males (1) (2) (3) (4)Female RR/Male RR -0.053*** -0.057*** -0.025*** -0.045***(0.017) (0.019) (0.009) (0.014)British Ruled 0.016 -0.025***(0.010) (0.009)Proportion Non-Landlord -0.039**(0.018)Demographic Controls Yes Yes Yes YesR-squared 0.087 0.087 0.101 0.092Observations 274,556 274,556 274,556 180,471Geo Controls Yes Yes Yes YesState Fixed Effects No No Yes NoYear of British Land Revenue Control No No No YesNote: Data source is Banerjee and Iyer (2005) for the topmost panel where observations are at the district level forthe years 1961, 1971, 1981 and 1991. Data source is National Sample Survey Organization (2000) for the second andthird panels where the dependent variable is a dummy variable that takes the value 1 if the respondent is illiterateand is 0 otherwise. Observations are at the individual level for the year 2000. Robust standard errors clustered atthe district level are in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% level ofsignificance respectively. Female RR/Male RR refers to the proportion of female to male rent receivers in a districtin 1921 and is our explanatory variable of interest. British Ruled refers to a dummy variable that assumes the value1 if the district was administered directly by the British and is 0 otherwise. Geo Controls include altitude, latitude,whether the district has black, red or alluvial soil, if the district is coastal and annual average rainfall. Altituderefers to the average altitude of the district in metres, Latitude refers to latitude of the centre of the district/districtheadquarters in degrees. Black soil, alluvial soil and red soil indicate dummy variables that assume the value 1 ifthe district has black soil, alluvial soil and red soil respectively; assuming the value 0 otherwise. Coastal district isa dummy variable that assumes the value 1 if the district is a coastal one and is 0 otherwise. Demographic controlsinclude dummies for the individual’s caste/tribe, religion and whether he/she resides in rural area and his/her age.Column (4) is identical to specifications in Banerjee and Iyer (2005).1344.8. TablesTable 4.6: Health OutcomesDistrict Prenatal No Child NotInfant Care From Professional WeighedMortality Doctor At Birth at BirthFemale RR/Male RR -4.750*** 0.038* -0.069*** -0.071***(0.808) (0.023) (0.026) (0.024)R-squared 0.617 0.295 0.275 0.333Observations 266 22379 22427 22389Got Got Saw AbdomenTetanus Folic Doctor ExaminedPrenatal Tablets Postnatal PostnatalFemale RR/Male RR -0.007 0.020 0.045* 0.040**(0.025) (0.025) (0.026) (0.018)R-squared 0.137 0.251 0.267 0.247Observations 22429 22425 22424 22409After Giving Birth:Advice on: Family Breast- Baby Child GotPlanning feeding Care Vitamin AFemale RR/Male RR 0.040** 0.059*** 0.046** 0.040**(0.019) (0.022) (0.022) (0.018)R-squared 0.166 0.238 0.201 0.098Observations 22409 22410 22409 21376Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Each cell is a separate regression. Regression specification is Column (3) of Table 4.5. Datasource is Banerjee and Iyer (2005)for the first column in the uppermost panel where observationsare at the district level for the year 1991. Here, robust standard errors clustered at the state levelare in parentheses. For all other columns in the table, data source is Demographic and HealthSurveys (1998-99). The dependent variables are dummy variables that take the value 1 if thesituation described in the variable is true and is 0 otherwise. Observations are at the individuallevel. Robust standard errors clustered at the district level are in parentheses. ***. ** and *indicate statistical significance at the 1%, 5% and 10% level of significance respectively. Also,demographic controls are included for individual level outcomes and they include dummies for theindividual’s caste/tribe, religion and whether he/she resides in rural area and his/her age. ***,** and * indicate statistical significance at the 1%, 5% and 10% level of significance respectively.All specifications also control for a dummy for whether the district was ruled by the British in thepast.1354.8. TablesTable 4.7: State Level Reported Crimes and ArrestsWomen are TotalVictims Crimes Rape Kidnap MurdersFemale RR/Male RR 2.537*** 1.955** 0.651 1.522**(0.596) (0.880) (1.209) (0.637)R-squared 0.828 0.704 0.751 0.557Observations 368 368 320 128Women are Men’s Murders of Property Crimes AgainstNot Victims Kidnaps Men Crimes Public OrderFemale RR/Male RR -3.630*** 1.391 2.605*** 1.607(0.993) (1.574) (0.582) (3.681)R-squared 0.582 0.524 0.759 0.497Observations 310 128 368 368Arrests for: All Crimes Non-Women Men’sAgainst Women Rape Crimes KidnapsFemale RR/Male RR 2.014*** 1.738* 2.134*** -5.334***(0.666) (0.859) (0.379) (0.773)R-squared 0.803 0.689 0.716 0.658Observations 320 320 320 293Chargesheeting Crimes AllRate Against Women CrimesFemale RR/Male RR 24.60*** -14.17(2.845) (10.40)R-squared 0.700 0.768Observations 272 272Geo Controls Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesRegion Fixed Effects Yes Yes Yes YesNote: Data source is Iyer et al(2012). Observations are at the state level. Robust standard errors clusteredat the state level are in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10%level of significance respectively. Specification is columns (3) of Table 4.5. Explanatory variables havebeen aggregated up to the state level from the district level. Crimes against women (men) are expressedas log of crimes per 1000 women (men). Property crimes and arrests are coded as log of crimes and arrestsrespectively per 1000 population.1364.8. TablesTable 4.8: Women’s Labour Market ParticipationDistrict Level: Female Female Male MaleWorkers Main Marginal Main MarginalFemale RR/Male RR 0.019*** 0.001 0.005 -0.003(0.004) (0.003) (0.004) (0.003)R-squared 0.772 0.552 0.655 0.639Observations 382 382 382 382Woman is: Currently Working for Earns Employed LastWorking Someone Else Cash 12 MonthsFemale RR/Male RR 0.054** 0.039*** 0.051*** 0.054**(0.024) (0.014) (0.018) (0.024)Demographic Controls Yes Yes Yes YesR-squared 0.175 0.109 0.109 0.177Observations 70176 70187 70187 70187Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Regression specification is Column (3) of Table 4.5. Data source is Primary Census Abstract, Censusof India (2011) for the upper panel where observations are at the district level for the year 2011 and isbased on the rural population. Robust standard errors clustered at the state level are in parentheses. Totalworkers comprise of main and marginal workers. Main (marginal) workers are individuals who worked formore (less) than 180 days preceding the census. The dependent variables are expressed as fraction of therural population. Data source is Demographic and Health Surveys (1998-99) for the lower panel whereobservations are at the individual level for women only. The dependent variables are dummy variablesthat take the value 1 if the situation described in the variable is true and is 0 otherwise. Robust standarderrors clustered at the district level are in parentheses. Demographic controls include dummies for theindividual’s caste/tribe, religion and whether he/she resides in rural area and his/her age.***, ** and *indicate statistical significance at the 1%, 5% and 10% level of significance respectively.Table 4.9: Measures of Women’s AutonomyVariable What to Obtain Purchase Stay WithCook Healthcare Jewellery FamilyFemale RR/Male RR -0.015 0.054* 0.023* 0.028*(0.014) (0.028) (0.012) (0.015)R-squared 0.102 0.101 0.053 0.064Observations 70181 70179 70177 70159Variable Go to Visit Spend Her Beaten SinceMarket Relatives/Friends Wages Age 15Female RR/Male RR 0.028 0.043** 0.028*** -0.018(0.021) (0.022) (0.010) (0.013)R-squared 0.207 0.135 0.038 0.052Observations 70171 70172 70187 70182Geo Controls Yes Yes Yes YesDemographic Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Regression specification is Column (3) of Table 4.5. Data source is Demographic and Health Surveys(1998-99). The dependent variables are dummy variables that take the value 1 if the situation describedin the variable is true and is 0 otherwise. Observations are at the individual level. Robust standard errorsclustered at the district level are in parentheses. ***, ** and * indicate statistical significance at the 1%,5% and 10% level of significance respectively. Demographic controls include dummies for the individual’scaste/tribe, religion and whether he/she resides in rural area and his/her age.1374.8. TablesTable 4.10: Mechanisms: Political Variables, Villages with Public Goodsand Health Seeking BehaviourPanel A: Fraction Won by Primary High DispensaryWomen Politicians School SchoolFemale RR/Male RR 0.032** -0.010** 0.003 0.003**(0.013) (0.004) (0.002) (0.001)R-squared 0.355 0.607 0.565 0.576Observations 345 315 315 315Geo Controls Yes Yes Yes YesRegion Fixed Effects YesState Fixed Effects Yes Yes YesPanel B: Health Tap Power for Giving BirthCentre Domestic Uses at HomeFemale RR/Male RR 0.0005* 0.040*** 0.01* -0.073**(0.0002) (0.005) (0.004) (0.036)R-squared 0.437 0.650 0.785 0.276Observations 315 315 315 3043Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Each cell is a separate regression. Data source is Iyer et al(2012) for the first column of Panel A; DHS(1998-99) for the fourth column in Panel B and Banerjee-Iyer (2005) for all other columns. Observations are at thestate level for the first column of Panel A; individual level for fourth column of Panel B and at the district level forall other columns. Robust standard errors clustered at the state level are in parentheses for all columns except thefourth column of Panel B, for which robust standard errors are clustered at the district level. The outcome variablescorresponding to public goods refer to the fraction of villages in a district having the public good in question. Theoutcome variable in the fourth column of Panel B is binary, which assumes the value 1 if the description in thevariable is true and is 0 otherwise. Explanatory variables have been aggregated up to the state level from the districtlevel first column of Panel A. The fourth column of Panel B also controls for the woman’s caste,religion, age and ifshe resides in rural area. ***, ** and * indicate statistical significance at the 1%, 5% and 10% level of significancerespectively. For other variable definitions, see Table Notes of Table 4.5.1384.8. TablesTable 4.11: Mechanisms: Log of Agricultural Yields at the District Level,Rice vs WheatVariable 15 Major Crops Rice WheatPanel A:Female RR/Male RR 0.032** 0.023** -0.017*(0.015) (0.011) (0.010)British Ruled 0.075 0.086 -0.032(0.060) (0.052) (0.039)R-squared 0.503 0.475 0.618Observations 8476 8044 7482Variable 15 Major Crops Rice WheatPanel B:Female RR/Male RR 0.036*** 0.0310*** -0.012(0.013) (0.010) (0.010)British Ruled 0.020 0.066 -0.058*(0.047) (0.044) (0.033)Prop Gross Cropped Irrigated 0.934*** 0.422*** 0.354***(0.142) (0.083) (0.095)Fertilizer Use (kg/hectare) 0.004*** 0.005*** 0.001*(0.001) (0.001) (0.001)Prop Crop Area Under HYV Variety Yes Yes YesR-squared 0.605 0.518 0.622Observations 7124 6692 5985Geo Controls Yes Yes YesYear Fixed Effects Yes Yes YesState Fixed Effects Yes Yes YesNote: Data source is Banerjee and Iyer (2005). Observations are at the district level for the years 1956-1987.Robust standard errors clustered at the district level are in parentheses. ***. ** and * indicate statisticalsignificance at the 1%, 5% and 10% level of significance respectively. For other variable definitions, seeTable Notes of Table 4.5.139Chapter 5ConclusionThis thesis has studied the situation of women and Scheduled Tribes inIndia.Chapter 2 has studied the impact of India’s National Rural EmploymentGuarantee Programme (NREGA) on consumption expenditure, householdfood security and individual time-use. Greater number of days worked bythe household and in some situations particularly those by women relativeto men, is found to increase spending on foods that can raise children’s nu-tritional status, improve food security and increase investment in educationof girls in the form of their school uniforms and fees. Also, women’s engage-ment in domestic tasks as their major activity is found to decline on accountof the programme. Time spent in school by younger girls is also found toincrease but no significant impact is found on the time spent performing do-mestic chores by children. Therefore contrary to social norms about genderroles, it is comforting to find that girls are unlikely to substitute for adultsin performing domestic tasks when adults work outside their homes.Chapter 3 compares health outcomes between Scheduled Tribes (STs)and Scheduled Castes (SCs) in India. Both these social groups have faceddisadvantage in the Indian society albeit for different historical reasons.However, they have been frequently grouped together as a single disad-vantaged group, even subjected to similar affirmative action policies andrarely been studied separately from each other. The chapter finds that STsperform poorly not only relative to the higher castes, but even relative tothe SCs in nearly all aspects of women’s and children’s health. It is foundthat relative to demand for healthcare that is likely to be influenced by one’seducation, exposure to the media, access to basic household facilities andwomen’s social status, the lack of availability of health services can largely140Chapter 5. Conclusionexplain the health disparity between the STs and SCs. This chapter, there-fore, attempts to argue for the need to study STs in isolation from the SCs.Further, the National Rural Health Mission implemented by the governmentsince 2005 is not found to decrease the shortage in the availability of publichealth infrastructure in rural, tribal areas.Chapter 4 has studied whether historical property rights for women haslong-term implications for current development outcomes. It is found thathigher relative female landownership in the past is associated with higherliteracy rate, lower infant mortality, better healthcare for, higher labourmarket participation of, greater reporting of crimes and arrest for crimesagainst and autonomy of women. Greater women’s political representationand role in agriculture as well as investment in public goods are found tobe likely explanations for how historical female property rights can influ-ence current outcomes, particularly those that affect women’s well-being inmodern India.141Bibliography[1] Acemoglu, Daron, James A. Robinson and Simon Johnson. 2001. “TheColonial Origins of Comparative Development: An Empirical Investi-gation.” American Economic Review 91: 1369-1401.[2] Acemoglu, Daron and Simon Johnson. 2005. “Unbundling Institu-tions.” Journal of Political Economy 113(5): 949-995.[3] Afridi, Farzana, Abhiroop Mukhopadhyay and Soham Sahoo. 2014.“Female Labour Force Participation and Child Education in In-dia: Evidence from the National Rural Employment GuaranteeScheme.”mimeo.[4] Afridi, Farzana, Vegard Iversen and M.R.Sharan. 2014. “Women Polit-ical Leaders, Corruption and Learning: Evidence from a Large PublicProgram in India.” International Growth Centre Working Paper.[5] Agarwal, Bina. 2002. “ “Bargaining” and Legal Change: Toward Gen-der Equality in India’s Inheritance Laws” IDS Working Paper.[6] Alesina, Alberto, Paola Giuliano and Nathan Nunn. 2013. “On theOrigin of Gender Roles: Women and the Plough.” The Quarterly Jour-nal of Economics 128(2): 469-530.[7] Anderson, Siwan and Mukesh Eswaran. 2009. “Determinants of Fe-male Autonomy: Evidence from Bangladesh.” Journal of DevelopmentEconomics 90(2): 179-191.[8] Azam, Mehtabul. 2012. “The Impact of Indian Job Guarantee Schemeon Labor Market Outcomes: Evidence from a Natural Experiment.”IZA Discussion Paper.142Bibliography[9] Banerjee, Abhijit and Lakshmi Iyer. 2005. “History, Institutions, andEconomic Performance: The Legacy of Colonial Land Tenure Systemsin India.” American Economic Review 95(4): 1190-1213.[10] Baviskar, Amita. 1994. “Fate of the Forest: Conservation and TribalRights.” Economic & Political Weekly 29(38): 2493-2501.[11] Baviskar, Amita. 1995. In the Belly of the River: Tribal ConflictsOver Development in the Narmada Valley, Second Edition. OxfordUniversity Press: New Delhi.[12] Baviskar, Amita. 1997. Displacement and the Bhilala Tribals of theNarmada Valley. In: Dreze, Jean, M.Samson and S.Singh (Eds.), TheDame and the Nation: Displacement and Resettlement in the NarmadaValley. Oxford University Press: New Delhi, 103-135.[13] Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, RohiniPande and Petia Topalova. 2009. “Powerful Women: Does ExposureReduce Bias?.” The Quarterly Journal of Economics 124(4): 1497-1540.[14] Beaman, Lori, Esther Duflo, Rohini Pande and Petia Topalova. 2012.“Female Leadership Raises Aspirations and Educational Attainmentfor Girls: A Policy Experiment in India.” Science 335: 582-586.[15] Beteille, Andre. 1986. “The Concept of Tribe with Special Referenceto India.” European Journal of Sociology 27: 310.[16] Bhalotra, Sonia R., and Irma Clots-Figueras. 2014. “Health and thePolitical Agency of Women.” American Economic Journal: EconomicPolicy 6(2): 164-197.[17] Bose, Nayana. 2015. “Raising Consumption through India’s NationalRural Employment Guarantee Scheme.” mimeo.[18] Clots-Figueras, Irma. 2012. “Are Female Leaders Good for Educa-tion? Evidence from India.” American Economic Journal: AppliedEconomics 4(1): 212-244.143Bibliography[19] Das, Subal and Kaushik Bose. 2015. “Adult Tribal Malnutrition inIndia: An Anthropometric and Socio-Demographic Review.” Anthro-pological Review 78(1): 47-65.[20] Das, Maitreyi Bordia, Soumya Kapoor and Denis Nikitin. 2010. “ACloser Look at Child Mortality among Adivasis.” World Bank PolicyResearch Working Paper 5231.[21] Desai, A.R.. 2005. Social Background of Indian Nationalism, PopularPrakashan.[22] Dreze, Jean and Reetika Khera. 2009. “The Battle for EmploymentGuarantee.”Frontline 26(1).[23] Dutta, Puja, Rinku Murgai, Martin Ravallion and Dominique Van DeWalle. 2012. “Does India’s Employment Guarantee Scheme GuaranteeEmployment?.”Economic and Political Weekly XLVII (16): 55-64.[24] Dyson, Tim and Mick Moore. 1983. “On Kinship Structure, FemaleAutonomy and Demographic Behaviour in India.” Population and De-velopment Review 9(1): 35-60.[25] Ekka, Alexius. 2000. “Jharkhand Tribals: Are They Really a Minor-ity?.” Economic & Political Weekly 35(52/53): 4610-4612.[26] Ekka, Alexius and Mohammad Asif. 2000. “Data Base onDevelopment-Induced-Displacement and Rehabilitation in Jharkhand1951-1995.” mimeo India Social Institute, New Delhi.[27] Eswaran, Mukesh, Bharat Ramaswami and Wilima Wadhwa. 2013.“Status, Caste and the Time Allocation of Women in Rural India.”Economic Development and Cultural Change 61(2): 311-333.[28] Gehrke, Esther. 2014. “An Employment Guarantee ad Risk Insur-ance? Assessing the Effects of NREGS on Agricultural ProductionDecisions.” BGPE Discussion Paper.144Bibliography[29] Guha, Ramachandra. 2007. “Adivasis, Naxalites and Indian Democ-racy.” Economic & Political Weekly 42(32): 3305-3312.[30] Hnatkovska, Viktoria, Amartya Lahiri and Sourabh Paul. 2012.“Castes and Labor Mobility.” American Economic Journal: AppliedEconomics 4(2): 274-307.[31] Hnatkovska, Viktoria, Amartya Lahiri and Sourabh Paul. 2013.“Breaking the Caste Barrier: Intergenerational Mobility in India.”Journal of Human Resources 48: 274-307.[32] Imbert, Clement and John Papp. 2015. “Labour Market Effects of So-cial Programs: Evidence from India’s Employment Guarantee,” Amer-ican Economic Journal: Applied Economics 7(2): 233-263.[33] Islam, Mahnaz and Anitha Sivasankaran. 2014. “How does Child La-bor respond to changes in Adult Work Opportunities? Evidence fromNREGA.” mimeo.[34] Iyer, Lakshmi. 2010. “Direct versus Indirect Colonial Rule in India:Long-term Consequences.” Review of Economics and Statistics 92(4):693-713.[35] Iyer, Lakshmi, Anandi Mani, Prachi Mishra, and Petia Topalova. 2012.“The Power of Political Voice: Women’s Political Representation andCrime in India.” American Economic Journal: Applied Economics4(4): 165-193.[36] Khera, Ritika and Nandini Nayak. 2009. “Women Workers and Percep-tions of the National Rural Employment Guarantee Act.” Economicand Political Weekly 44(43): 49-57.[37] Khosla, Rishabh. 2011. “Caste, Politics and Public Good Distributionin India: Evidence from NREGS in Andhra Pradesh.” Economic andPolitical Weekly XLVI(12): 63-69.145Bibliography[38] Kijima, Yoko. 2006. “Caste and Tribe Inequality: Evidence from India,1983-1999.” Economic Development and Cultural Change 54(2): 369-404.[39] Liu, Yanyan and Klaus Deininger. 2013. “Poverty Impacts of In-dia’s National Rural Employment Guarantee Scheme: Evidence fromAndhra Pradesh.” mimeo.[40] Mahajan, Kanika. 2015. “Farm Wages and Public Works: How Ro-bust are the Impacts of the National Rural Employment GuaranteeScheme?.” Indian Growth and Development Review 8(1): 19-72.[41] Maharatna, Arup. 2000. “Fertility, Mortality and Gender Bias AmongTribal Population: an Indian Perspective.” Social Science & Medicine50: 1333-1351.[42] Maharatna, Arup and Rasika Chikte. 2004. “Demography of TribalPopulation in Jharkhand 1951-1991.” Economic & Political Weekly39 (46/47): 5053-5062.[43] Maharatna, Arup. 2011. “How Can ‘Beautiful’ be ‘Backward’? Tribesof India in a Long-Term Demographic Perspective.” Economic & Po-litical Weekly XLVI (4): 42-52.[44] Maiorano, Diego. 2014. “The Politics of the Mahatma Gandhi Na-tional Rural Employment Guarantee Act in Andhra Pradesh.” WorldDevelopment 58: 95-105.[45] Mitra, Aparna. 2008. “The Status of Women Among the ScheduledTribes in India.” The Journal of Socio-Economics 37: 1202-1217.[46] Mukhopadhyay, Abhiroop. 2012. “Mahatma Gandhi National RuralEmployment Guarantee Scheme: Falling Demands or Funds Crunch?.”Ideas for India.[47] Mukhopadhyay, Abhiroop. 2012. “The Political Economy of MahatmaGandhi National Rural Employment Guarantee Scheme.” ESID Work-ing Paper No 15.146Bibliography[48] Niehaus, Paul and Sandip Sukhtankar. 2012. “The Marginal Rate ofCorruption in Public Programs: Evidence from India.” Journal ofPublic Economics 104: 52-64.[49] Qian, Nancy. 2008. ‘”Missing Women and the Price of Tea in China:The Effect of Sex-Specific Income on Sex Imbalance.” The QuarterlyJournal of Economics 123(3): 1251-1285.[50] Ramaiah, Avatthi. 2015. “Health Status of Dalits in India.” Economic& Political Weekly L (40): 70-74.[51] Ravi, Shamika and Monika Engler. 2015. “Workfare as an EffectiveWay to Fight Poverty: The Case of India’s NREGS.” World Develop-ment 67: 57-71.[52] Reddy, Narasimha D. 2012. “The Long Road to Social Security: TheNational Rural Employment Guarantee Scheme in Andhra Pradesh.”Paper 11, Centre for Development Studies, Trivandrum, Kerala, In-dia.[53] Sarkar, Sandip, Sunil Mishra, Harishwar Dayal and Dev Nathan. 2006.“Development and Deprivation of Scheduled Tribes.” Economic & Po-litical Weekly 41 (46): 4824-4827.[54] Singh, Huidrom, Suraj, Manisha Ghritlahre and Subal Das. 2014. “Nu-tritional Status among Females of Bhaina Tribe of Bilaspur, Chattis-garh, India: An Anthropological Insight.” Journal of Anthropology2014: 1-7.[55] Sundar, Nandini (Ed.). 2009. Legal Grounds: Natural Resources, Iden-tity and the Law of Jharkhand. Oxford University Press.[56] Usami, Yoshifumi. 2006. “Change in the Work-Force in British India,1881-1921.” XIV International Economic History Congress, Helsinki.[57] Xaxa, Virginius. 1999. “Tribes as Indigenous People of India.” Eco-nomic & Political Weekly 34 (51): 3589-3595.147[58] Xaxa, Virginius. 2001. “Protective Discrimination: Why ScheduledTribes Lag Behind Scheduled Castes.” Economic & Political Weekly36 (29): 2765-2772.[59] Zimmermann, Laura. 2015. “Why Guarantee Employment? Evidencefrom a Large Indian Public-Works Program.” University of GeorgiaWorking Paper.148Appendix AAppendix to Chapter 2A.1 Appendix Tables149A.1.AppendixTablesTable A.1: Women’s Labour Market Participation as % of Total Workers in 18 Major Indian StatesState % Cultivate % Agr Labour % HH Industry % Fem Work Rank 2001 Rank 2011Andhra Pradesh 14.01 58.00 5.22 36.2 10 7Bihar 15.27 60.77 6.83 19.1 26 24Chattisgarh 31.32 54.44 1.47 39.7 3 3Gujarat 17.78 47.14 1.97 23.4 16 20Haryana 32.78 13.60 3.59 17.8 18 27Himachal Pradesh 76.24 4.75 1.42 44.8 2 1Jharkhand 32.60 44.81 4.91 29.1 19 15Karnataka 19.03 40.33 4.94 31.9 13 12Kerala 3.89 14.68 3.42 18.2 32 25Madhya Pradesh 28.47 51.47 3.92 32.6 12 11Maharashtra 29.61 39.92 3.18 31.1 15 14Orissa 12.92 57.78 6.10 27.2 20 16Punjab 9.94 19.14 7.50 13.9 25 33Rajasthan 52.64 24.22 2.46 35.1 11 9Tamil Nadu 13.24 41.61 6.76 31.8 14 13Uttarakhand 64.00 8.84 3.41 26.7 17 17Uttar Pradesh 22.21 38.43 9.67 16.7 31 30West Bengal 7.67 34.03 16.69 18.1 28 26India 24.01 55.21 5.71 30.02Note: Data source is the Census of India, 2001 and 2011 from the publication“Statistical Profile on Women Labour” by the Labour Bureau, Ministry of Labour& Employment, Govt of India (2012-13). Columns 2-5 denote percentage of women workers in the respective categories with respect to total number of workerscorresponding to Census, 2011. Columns 6 and 7 indicate ranking of the state in India in terms of overall women’s labour market participation according toCensus, 2001 and Census, 2011 respectively. Andhra Pradesh refers to undivided Andhra Pradesh and the state we study in this paper.150A.1. Appendix TablesTable A.2: Some Evidence of Women’s Control over their NREGA Wages% Sampled Women Workers who:Collect Their Own Wages 79Keep Their Own Wages 68Earned Any Cash Income(Other than NREGA) 30Note: Data source is Dreze and Khera (2009) from their NREGA, 2008survey in 6 North Indian states-Bihar, Chattisgarh, Jharkhand, MadhyaPradesh, Rajasthan and Uttar Pradesh.Table A.3: Female Share of Employment in NREGA: % of Total Person-DaysState FY 2006-07 FY 2007-08 FY 2008-09 FY 2009-10 FY 2010-11Andhra Pradesh 55 58 58 58 57Bihar 17 28 30 30 28Chattisgarh 39 42 47 49 49Gujarat 50 47 43 48 44Haryana 31 34 31 35 36Himachal Pradesh 12 30 39 46 48Jharkhand 39 27 29 34 33Karnataka 51 50 50 37 46Kerala 66 71 85 88 90Madhya Pradesh 43 42 43 44 44Maharashtra 37 40 46 40 46Orissa 36 36 38 36 39Punjab 38 16 25 26 34Rajasthan 67 69 67 67 68Tamil Nadu 81 82 80 83 83Uttarakhand 30 43 37 40 40Uttar Pradesh 17 15 18 22 21West Bengal 18 17 27 33 34India 40 43 48 48 48Note: Data source is the “MGNREGA Sameeksha 2006-2012” published by the Ministry of Rural Development, Governmentof India. FY refers to financial year that ranges from April 1 of one year to March 31 of the following year. Andhra Pradeshreferes to undivided Andhra Pradesh and the state we study in this paper.151A.1.AppendixTablesTable A.4: Average NREGA Wage and Casual Wage in Rural IndiaState Average NREGA Average Casual Casual Wage Casual Wage Male-FemaleWage Wage Overall Males Females DifferenceAndhra Pradesh 91.9 98.5 115.4 75.7 39.7Bihar 97.5 79.4 81 65.8 15.2Chattisgarh 82.3 68.8 70.8 65.5 5.3Gujarat 89.3 83.3 87.3 71 16.3Haryana 150.9 139.6 146.1 99.1 47Himachal Pradesh 109.5 139.6 141.4 110.2 31.2Jharkhand 97.7 101.2 103.6 82.2 21.4Karnataka 86 84.5 96.9 62.8 34.1Kerala 120.6 206.5 226.6 119.3 107.3Madhya Pradesh 83.7 69 74.5 58.1 16.4Maharashtra 94.3 75.2 86 58.2 27.8Orissa 105.9 75.6 81 59.1 21.9Punjab 123.5 130.4 133.5 91.8 41.7Rajasthan 87.4 125.7 132.3 94.3 38Tamil Nadu 71.6 110.8 132.1 72.6 59.5Uttarakhand 99 118.7 122.1 96.7 25.4Uttar Pradesh 99.5 94.3 97 69.2 27.8West Bengal 90.4 85.3 87.8 65.9 21.9India 90.2 93.1 101.5 68.9 32.6Note: Wages are reported as Rs/day. Data source is the “MGNREGA Sameeksha 2006-2012” published by the Ministry ofRural Development, Government of India. FY refers to financial year that ranges from April 1 of one year to March 31 ofthe following year. Andhra Pradesh referes to undivided Andhra Pradesh and the state we study in this paper.152Appendix BAppendix to Chapter 3B.1 Appendix Tables153B.1. Appendix TablesTable B.1: Woman is UnderweightVariable (1) (2)If Scheduled Tribe 0.114*** 0.156***(0.014) (0.025)If Scheduled Caste 0.101*** 0.141***(0.008) (0.016)If OBC 0.053*** 0.074***(0.007) (0.013)Birth Cohort 2*ST -0.018(0.028)Birth Cohort 3*ST -0.014(0.027)Birth Cohort 4*ST -0.039(0.028)Birth Cohort 5*ST -0.067**(0.027)Birth Cohort 6*ST -0.076**(0.029)Birth Cohort 7*ST -0.084*(0.043)Birth Cohort 2*SC -0.006(0.022)Birth Cohort 3*SC -0.027(0.021)Birth Cohort 4*SC -0.043**(0.020)Birth Cohort 5*SC -0.053***(0.020)Birth Cohort 6*SC -0.071***(0.020)Birth Cohort 7*SC -0.085**(0.035)Birth Cohort 2*OBC -0.003(0.016)Birth Cohort 3*OBC -0.010(0.017)Birth Cohort 4*OBC -0.019(0.016)Birth Cohort 5*OBC -0.041**(0.016)Birth Cohort 6*OBC -0.040**(0.018)Birth Cohort 7*OBC -0.048(0.032)Birth Cohort Dummies No YesReligion/Rural Dummies Yes YesWoman’s Age Yes YesState Fixed Effects Yes YesObservations 83098 83098Note: Data source is DHS (1998-99). Robust standard errors clustered at the districtlevel are in parentheses. ***, ** and * indicate statistical significance at the 1% , 5% and10% level of significance respectively. Regressions weighted by survey weight. Omittedcategory is non-SC/ST/OBC social group. Religion dummies included control for whetherthe respondent is Hindu, Muslim, Christian or Sikh; the omitted category being all otherreligions. Seven (roughly) 5 year birth cohorts have been computed; starting from yearof birth 1948 upto 1985; later birth cohort numbers refer to younger women.154B.1. Appendix TablesTable B.2: Gender Differences in Child MortalityPanel A: Female Child Male ChildVariable Neonatal Mortality Neonatal MortalityIf Scheduled Tribe 0.006 0.006(0.004) (0.004)If Scheduled Caste 0.005* 0.006**(0.003) (0.002)If OBC 0.005** 0.007***(0.002) (0.002)Observations 88412 95675Panel B: Female Child Male ChildVariable Infant Mortality Infant MortalityIf Scheduled Tribe 0.002 0.015***(0.005) (0.004)If Scheduled Caste 0.013*** 0.014***(0.003) (0.004)If OBC 0.004 0.012***(0.003) (0.003)Observations 88412 95675Panel C: Female Child Male ChildVariable Child Mortality Child MortalityIf Scheduled Tribe 0.015** 0.025***(0.006) (0.005)If Scheduled Caste 0.024*** 0.024***(0.004) (0.004)If OBC 0.006* 0.015***(0.004) (0.004)Observations 88412 95675Child’s Year of Birth Fixed Effects Yes YesBirth Order Control Yes YesIf Child is of a Single Birth Yes YesMother’s Age at Birth Yes YesReligion/Rural Dummies Yes YesState Fixed Effects Yes YesNote: Data source is DHS (1998-99). Robust standard errors clustered at the district level are in parentheses. ***, ** and *indicate statistical significance at the 1% , 5% and 10% level of significance respectively. Sample restricted to all births atthe current place of residence of the mother.Regressions weighted by survey weight. Omitted category is non-SC/ST/OBCsocial group. Religion dummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh; the omittedcategory being all other religions.155B.1. Appendix TablesTable B.3: Gender Composition of a Woman’s ChildrenVariable Proportion Daughters to Daughters toDaughters Sons Sons AliveIf Scheduled Tribe 0.011** 0.107*** 0.093***(0.006) (0.018) (0.018)If Scheduled Caste 0.002 0.070*** 0.058***(0.004) (0.014) (0.014)If OBC 0.002 0.034*** 0.039***(0.003) (0.012) (0.012)Religion/Rural Dummies Yes Yes YesWoman’s Age Yes Yes YesState Fixed Effects Yes Yes YesObservations 80239 68854 66860Note: Data source is DHS (1998-99).Robust standard errors clustered at the district level are in parentheses.***, ** and * indicate statistical significance at the 1% , 5% and 10% level of significance respectively.Regressions weighted by survey weight. Omitted category is non-SC/ST/OBC social group. Religiondummies included control for whether the respondent is Hindu, Muslim, Christian or Sikh; the omittedcategory being all other religions.156Appendix CAppendix to Chapter 4C.1 Appendix TablesTable C.1: District Level Crimes per 1000 populationVariable Total Violent AttemptedCrimes Crimes Dacoity MurderFemale RR/Male RR 0.140*** 0.004 0.004 -12.08**(0.046) (0.004) (0.003) (4.737)R-squared 0.254 0.447 0.503 0.373Observations 552 552 549 218Variable Petty Stealing Women OtherCrimes Crimes Kidnaps KidnapsFemale RR/Male RR 0.008*** 0.099*** -4.310*** -0.966(0.002) (0.024) (1.085) (1.321)R-squared 0.220 0.336 0.398 0.145Observations 552 551 218 218Geo Controls Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Data source is Banerjee and Iyer(2005). Observations are at the district level.Robust standard errors clustered at the district level are in parentheses. ***. ** and *indicate statistical significance at the 1%, 5% and 10% level of significance respectively.Regression specification is column (3) of Table 4.5. For other variable definitions, seeTable Notes of Table 4.5.157C.1. Appendix TablesTable C.2: District Level Outcomes: Colonial Land Tenure SystemVariable Literacy InfantRate MortalityFemale RR/Male RR 0.020*** -4.694***(0.003) (1.402)Proportion Non-Landlord 0.057*** -33.34**(0.019) (11.98)R-squared 0.572 0.402Observations 583 165Variable Male Main Male Marginal Female Main Female MarginalWorkers Workers Workers WorkersFemale RR/Male RR 0.009 -0.007 0.019*** 0.003(0.006) (0.005) (0.003) (0.004)Proportion Non-Landlord 0.020** -0.025*** 0.008 -0.023(0.008) (0.007) (0.007) (0.015)R-squared 0.494 0.414 0.723 0.193Observations 225 225 225 225Variable Total Violent AttemptedCrimes Crimes Dacoity MurderFemale RR/Male RR 0.195*** 0.010 0.006 -11.25***(0.039) (0.010) (0.009) (2.931)Proportion Non-Landlord -0.748*** -0.066*** -0.037** -52.92***(0.235) (0.023) (0.018) (14.45)R-squared 0.156 0.180 0.231 0.165Observations 549 549 546 217Variable Petty Stealing Women OtherCrimes Crimes Kidnaps KidnapsFemale RR/Male RR 0.010*** 0.146*** -2.589** -1.372*(0.002) (0.025) (1.114) (0.789)Proportion Non-Landlord -0.001 -0.311*** -17.36*** -13.53**(0.008) (0.108) (4.863) (6.141)R-squared 0.134 0.205 0.114 0.085Observations 549 548 217 217Geo Controls Yes Yes Yes YesYear of British Land Revenue Control Yes Yes Yes YesNote: Observations are at the district level. Each cell is a separate regression and is identical to Banerjee and Iyer (2005)specifications. See Table Notes for tables with the corresponding outcome variables for details about data sources, variabledefinitions, controls, the inclusion of year fixed effects and the level of clustering of the standard errors. No state fixed effectsare included as some states do not have variation in historic proportion non-landlord that is taken from Banerjee and Iyer(2005).158C.1. Appendix TablesTable C.3: State Level Outcomes: Colonial Land Tenure SystemWomen are TotalVictims Crimes Rape Kidnap MurdersFemale RR/Male RR 3.931*** 2.994* 1.376 0.294(0.648) (1.479) (0.845) (0.577)Average Non-Landlord -1.287*** -1.727** -1.245** 0.183(0.194) (0.636) (0.449) (0.391)R-squared 0.823 0.655 0.740 0.493Observations 345 345 300 120Women are Men’s Murders of Property Crimes AgainstNot Victims Kidnaps Men Crimes Public OrderFemale RR/Male RR -6.215*** -1.461* 2.197* -0.111(0.508) (0.696) (1.189) (3.497)Average Non-Landlord 1.808*** 0.396 -0.738* -1.477(0.265) (0.452) (0.370) (1.122)R-squared 0.650 0.703 0.657 0.567Observations 290 120 345 345Arrests for: All Crimes Non-Women Men’sAgainst Women Rape Crimes KidnapsFemale RR/Male RR 2.286** 2.819* 4.943*** -6.813***(0.862) (1.520) (0.325) (0.744)Average Non-Landlord -0.783** -1.435* -1.316*** 1.757***(0.286) (0.773) (0.110) (0.191)R-squared 0.798 0.598 0.686 0.635Observations 300 300 300 276Chargesheeting Crimes AllRate Against Women CrimesFemale RR/Male RR 31.84** 2.233(14.59) (14.86)Average Non-Landlord -10.28** -4.477(4.566) (5.074)R-squared 0.634 0.745Observations 255 255Geo Controls Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesYear of British Land Revenue Control Yes Yes Yes YesNote: Observations are at the state level. Each cell is a separate regression and is identical to Banerjee and Iyer (2005)specification. See Table Notes for tables with the corresponding outcome variables for details about data source, variabledefinitions, controls and the level of clustering of the standard errors. No region fixed effects are included. Banerjee andIyer (2005) colonial proportion non-landlord at the district level have been aggregated up to the state level.159C.1. Appendix TablesTable C.4: Woman’s Labour Market Participation: Rural vs UrbanVariable Currently Working Works For Someone ElsePanel A: Rural SampleFemale RR/Male RR 0.058** 0.040***(0.026) (0.014)British Rule Dummy -0.048*** -0.011(0.016) (0.013)R-squared 0.210 0.143Observations 50782 50791Panel B: Urban SampleFemale RR/Male RR 0.043 0.016(0.033) (0.025)British Rule Dummy 0.018 0.007(0.017) (0.015)R-squared 0.062 0.059Observations 19394 19396Variable Earns Cash Employed Last 12 MonthsPanel C: Rural SampleFemale RR/Male RR 0.056*** 0.056**(0.019) (0.024)British Rule Dummy -0.015 -0.043***(0.014) (0.016)R-squared 0.150 0.211Observations 50791 50791Panel D: Urban SampleFemale RR/Male RR 0.032 0.051(0.029) (0.033)British Rule Dummy 0.017 0.020(0.015) (0.017)R-squared 0.055 0.062Observations 19396 19396Geo Controls Yes YesDemographic Controls Yes YesState Fixed Effects Yes YesNote: Data source is Demographic and Health Surveys (1998-99). The dependent variables aredummy variables that take the value 1 if the situation described in the variable is true and is0 otherwise. Observations are at the individual level. Robust standard errors clustered at thedistrict level are in parentheses. ***. ** and * indicate statistical significance at the 1%, 5% and10% level of significance respectively. Demographic controls include dummies for the individual’scaste/tribe, religion and his/her age. For other variable definitions, see Table Notes of Table 4.5.160C.1. Appendix TablesTable C.5: Alternative Measures of Women’s AutonomyVariable Autonomy Index 1 Autonomy Index 2Female RR/Male RR 0.090 0.082**(0.055) (0.041)British Rule Dummy 0.014 0.025(0.025) (0.018)R-squared 0.130 0.106Observations 70153 70156Variable Autonomy Index 3 Autonomy Index 4Female RR/Male RR 0.162* 0.153**(0.091) (0.078)British Rule Dummy 0.018 0.029(0.041) (0.036)R-squared 0.196 0.197Observations 70138 70141Geo Controls Yes YesDemographic Controls Yes YesState Fixed Effects Yes YesNote: Data source is Demographic and Health Surveys (1998-99). Observations are at the individual level.Robust standard errors clustered at the district level are in parentheses. ***. ** and * indicate statisticalsignificance at the 1%, 5% and 10% level of significance respectively. Demographic controls include dummiesfor the individual’s caste/tribe, religion and whether he/she resides in rural area and his/her age. Fordefinitions of the dependent variables for Panels A and B, see Table Notes of Table 10. Autonomy Index 3includes the components of Autonomy Index 1 along with 1 point each if the respondent needs no permissionto go to market, visit relatives/friends and is allowed to have money set aside. Autonomy Index 4 includescomponents of Autonomy Index 2 along with 1 point each if the respondent needs no permission to go tomarket, visit relatives/friends and is allowed to have money set aside. The maximum value for AutonomyIndex 3 (4) is 8 (6) and minimum being 0 (0). The higher the value of the autonomy indices, the greateris woman’s autonomy. For other variable definitions, see Table Notes of Table 4.5.161C.1. Appendix TablesTable C.6: Mechanisms: Log of Agricultural Yields at the District Level,Rice vs WheatVariable 15 Major Crops Rice WheatPanel A:Female RR/Male RR 0.018* -0.003 -0.029**(0.011) (0.014) (0.013)Proportion Non-Landlord 0.160** 0.171** 0.215***(0.072) (0.082) (0.069)R-squared 0.431 0.407 0.524Observations 5279 5261 4452Variable 15 Major Crops Rice WheatPanel B:Female RR/Male RR 0.034*** 0.015 -0.020**(0.008) (0.011) (0.010)Proportion Non-Landlord 0.043 0.077 0.087(0.054) (0.068) (0.064)Prop Gross Cropped Irrigated 0.648*** 0.390*** 0.439***(0.118) (0.102) (0.117)Fertilizer Use (kg/hectare) 0.006*** 0.005*** 0.001(0.001) (0.001) (0.001)Prop Crop Area Under HYV Variety Yes Yes YesR-squared 0.599 0.490 0.566Observations 4428 4387 3557Geo Controls Yes Yes YesYear Fixed Effects Yes Yes YesYear of British Land Revenue Control Yes Yes YesNote: Data source is Banerjee and Iyer (2005). Observations are at the district level for the years 1956-1987. Robust standard errors clustered at the district level are in parentheses. ***. ** and * indicatestatistical significance at the 1%, 5% and 10% level of significance respectively. The specification in thistable is analogous to Banerjee and Iyer (2005) specification. For other variable definitions, see Table Notesof Table 4.5.162C.1. Appendix TablesTable C.7: Alternative Specification of Female Property Rights IVariable Literacy Woman is Man is InfantRate Illiterate Illiterate MortalityFemale RR/Total RR 0.123** -0.072* -0.074** -11.19(0.058) (0.041) (0.035) (17.64)R-squared 0.703 0.177 0.103 0.607Observations 973 251,750 283,976 266Woman Had: Prenatal Care No Professional Child Not GotFrom Doctor at Birth Weighed at Birth TetanusFemale RR/Total RR 0.102* -0.164** -0.074** -0.040(0.061) (0.069) (0.035) (0.070)R-squared 0.295 0.274 0.333 0.137Observations 22379 22427 22389 22429Woman Had: Got Folic Saw Doctor Abdomen ExaminedTablets Postnatal PostnatalFemale RR/Total RR 0.041 0.117* 0.106**(0.064) (0.062) (0.044)R-squared 0.251 0.267 0.247Observations 22425 22424 22409After Birth: Family Planning Breastfeeding Babycare Child GivenAdvice Advice Advice Vitamin AFemale RR/Total RR 0.114*** 0.157*** 0.123** 0.100**(0.039) (0.052) (0.053) (0.050)R-squared 0.166 0.238 0.201 0.098Observations 22409 22410 22409 21376Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Each cell is a separate regression and is identical to the specification in Column (3) of Table 4.5. See Table Notesin the main text of Chapter 4 for tables with the corresponding outcome variables for details about data sources, variabledefinitions, controls, the inclusion of year fixed effects and the level of clustering of the standard errors.163C.1. Appendix TablesTable C.8: Alternative Specification of Female Property Rights IIVariable Female Main Female Marginal Male Main Male MarginalWorkers Workers Workers WorkersFemale RR/Total RR 0.061*** 0.004 0.028* -0.018(0.014) (0.011) (0.016) (0.012)R-squared 0.772 0.552 0.657 0.641Observations 382 382 382 382Woman is: Currently Working for Earning Employed forWorking Someone Else Cash Last 12 MonthsFemale RR/Total RR 0.153** 0.102** 0.135*** 0.146**(0.071) (0.042) (0.052) (0.072)R-squared 0.175 0.109 0.109 0.177Observations 70176 70187 70187 70187Woman Can What to To Obtain Purchase Stay withDecide: Cook Healthcare Jewellery FamilyFemale RR/Total RR -0.013 0.172*** 0.067** 0.073*(0.047) (0.058) (0.028) (0.040)R-squared 0.102 0.101 0.053 0.064Observations 70181 70179 70177 70159Woman Can/ Go to Visit Friends/ Spend Her Been BeatenHas: Market Relatives Wages Since Age 15Female RR/Total RR 0.065 0.123** 0.086*** -0.039(0.056) (0.053) (0.023) (0.041)R-squared 0.207 0.135 0.039 0.052Observations 70171 70172 70187 70182Geo Controls Yes Yes Yes YesState Fixed Effects Yes Yes Yes YesNote: Each cell is a separate regression and is identical to the specification in Column (3) of Table 4.5. See Table Notesin the main text of Chapter 4 for tables with the corresponding outcome variables for details about data sources, variabledefinitions, controls, the inclusion of year fixed effects and the level of clustering of the standard errors.164C.1. Appendix TablesTable C.9: Alternative Specification of Female Property Rights IIIWomen are TotalVictims Crimes Rape Kidnap MurdersFemale RR/Total RR 4.438*** 3.255* 1.225 2.742**(1.117) (1.624) (2.055) (1.077)R-squared 0.827 0.699 0.752 0.558Observations 368 368 320 128Women are Men’s Murders of Property Crimes AgainstNot Victims Kidnaps Men Crimes Public OrderFemale RR/Total RR -6.688*** 2.273 4.517*** 2.267(1.696) (2.803) (1.143) (6.612)R-squared 0.587 0.519 0.752 0.494Observations 310 128 368 368Arrests for: All Crimes Non-Women Men’sAgainst Women Rape Crimes KidnapsFemale RR/Total RR 3.525** 2.909* 3.931*** -9.728***(1.209) (1.512) (0.582) (1.220)R-squared 0.802 0.685 0.719 0.663Observations 320 320 320 293Chargesheeting Crimes AllRate Against Women CrimesFemale RR/Total RR 42.88*** -27.63(6.201) (17.19)R-squared 0.697 0.770Observations 272 272Geo Controls Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesRegion Fixed Effects Yes Yes Yes YesNote: Each cell is a separate regression and is identical to the specification in Column (3) of Table 4.5.See Table Notes in the main text of Chapter 4 for tables with the corresponding outcome variables fordetails about data sources, variable definitions, controls, the inclusion of year fixed effects and the level ofclustering of the standard errors.165

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0307290/manifest

Comment

Related Items